Search results for: imbalanced multiclass classifi-cation

Authors: Konstantinos Liagkouras, Konstantinos Metaxiotis

Abstract:

In this paper we present a classification of the various technologies applied for the solution of the portfolio selection problem according to the discipline and the methodological framework followed. We provide a concise presentation of the emerged categories and we are trying to identify which methods considered obsolete and which lie at the heart of the debate. On top of that, we provide a comparative study of the different technologies applied for efficient portfolio construction and we suggest potential paths for future work that lie at the intersection of the presented techniques.

Keywords: portfolio selection, optimization techniques, financial models, stochastic, heuristics

Procedia PDF Downloads 432

Authors: Elizabeth Malebogo Mosepele

Abstract:

Road transport has become increasingly common in the world; people rely on road networks for transportation purpose on a daily basis. However, environmental impact of roads on surrounding landscapes extends their potential effects even further. This study investigates the impact of road network on natural vegetation. The study will provide baseline knowledge regarding roadside vegetation and would be helpful in future for conservation of biodiversity along the road verges and improvements of road verges. The general hypothesis of this study is that the amount and condition of road side vegetation could be explained by road network conditions. Remote sensing techniques were used to analyze vegetation conditions. Landsat 8 OLI image was used to assess vegetation cover condition. NDVI image was generated and used as a base from which land cover classes were extracted, comprising four categories viz. healthy vegetation, degraded vegetation, bare surface, and water. The classification of the image was achieved using the supervised classification technique. Road networks were digitized from Google Earth. For observed data, transect based quadrats of 50*50 m were conducted next to road segments for vegetation assessment. Vegetation condition was related to road network, with the multinomial logistic regression confirming a significant relationship between vegetation condition and road network. The null hypothesis formulated was that 'there is no variation in vegetation condition as we move away from the road.' Analysis of vegetation condition revealed degraded vegetation within close proximity of a road segment and healthy vegetation as the distance increase away from the road. The Chi Squared value was compared with critical value of 3.84, at the significance level of 0.05 to determine the significance of relationship. Given that the Chi squared value was 395, 5004, the null hypothesis was therefore rejected; there is significant variation in vegetation the distance increases away from the road. The conclusion is that the road network plays an important role in the condition of vegetation.

Keywords: Chi squared, geographic information system, multinomial logistic regression, remote sensing, road side vegetation

Procedia PDF Downloads 432

Authors: Jeonghwan Jeon

Abstract:

With the advent of IPTV in the fierce competition with existing broadcasting system, it is emerged as an important issue to predict how much the adoption of IPTV service will be. This paper aims to suggest a prediction model for adopting IPTV using classification and Ranking Belief Simplex (CaRBS). A simplex plot method of representing data allows a clear visual representation to the degree of interaction of the support from the variables to the prediction of the objects. CaRBS is applied to the survey data on the IPTV adoption.

Keywords: prediction, adoption, IPTV, CaRBS

Procedia PDF Downloads 412

Authors: Lydia Novozhilova, Vladimir Urazhdin

Abstract:

An optimization problem for reinforced thin-walled vessels under uniform external pressure is considered. The conventional approaches to optimization generally start with pre-defined geometric parameters of the vessels, and then employ analytic or numeric calculations and/or experimental testing to verify functionality, such as stability under the projected conditions. The proposed approach consists of two steps. First, the feasibility domain will be identified in the multidimensional parameter space. Every point in the feasibility domain defines a design satisfying both geometric and functional constraints. Second, an objective function defined in this domain is formulated and optimized. The broader applicability of the suggested methodology is maximized by implementing the Support Vector Machines (SVM) classification algorithm of machine learning for identification of the feasible design region. Training data for SVM classifier is obtained using the Simulation package of SOLIDWORKS®. Based on the data, the SVM algorithm produces a curvilinear boundary separating admissible and not admissible sets of design parameters with maximal margins. Then optimization of the vessel parameters in the feasibility domain is performed using the standard algorithms for the constrained optimization. As an example, optimization of a ring-stiffened closed cylindrical thin-walled vessel with semi-spherical caps under high external pressure is implemented. As a functional constraint, von Mises stress criterion is used but any other stability constraint admitting mathematical formulation can be incorporated into the proposed approach. Suggested methodology has a good potential for reducing design time for finding optimal parameters of thin-walled vessels under uniform external pressure.

Keywords: design parameters, feasibility domain, von Mises stress criterion, Support Vector Machine (SVM) classifier

Procedia PDF Downloads 327

Authors: R. Gwynn, P. Panwalkar, K. Veravalli , M. Tofighi, R. Clement, A. Mofidi

Abstract:

The aim of this study was to access the impact of Covid-19 and lockdown on the incidence, injury pattern, and treatment of tibial plateau fractures in a combined rural and urban population in wales. Methods: Retrospective study was performed to identify tibial plateau fractures in 15-month period of Covid-19 lockdown 15-month period immediately before lockdown. Patient demographics, injury mechanism, injury severity (based on Schatzker classification), and associated injuries, treatment methods, and outcome of fractures in the Covid-19 period was studied. Results: The incidence oftibial plateau fracture was 9 per 100000 during Covid-19, and 8.5 per 100000, and both were similar to previous studies. The average age was 52, and female to male ratio was 1:1 in both control and study group. High energy injury was seen in only 20% of the patients and 35% in the control groups (2=12, p<0025). 14% of the covid-19 population sustained other injuries as opposed 16% in the control group(2=0.09, p>0.95). Lower severity isolated lateral condyle fracturesinjury (Schatzker 1-3) were seen in 40% of fractures this was 60% in the control populations. Higher bicondylar and shaft fractures (Schatzker 5-6) were seen in 60% of the Covid-19 group and 35% in the control groups(2=7.8, p<0.02). Treatment mode was not impacted by Covid-19. The complication rate was low in spite of higher number of complex fractures and the impact of covid-19 pandemic. Conclusion: The associated injuries were similar in spite of a significantly lower mechanism of injury. There were unexpectedly worst tibial plateau fracture based Schatzker classification in the Covid-19 period as compared to the control groups. This was especially relevant for medial condyle and shaft fractures. This was postulated to be caused by reduction in bone density caused by lack of vitamin D and reduction in activity. The treatment mode and outcome was not impacted by the impact of Covid-19 on care for tibial plateau fractures.

Keywords: Covid-19, knee, tibial plateau fracture, trauma

Procedia PDF Downloads 125

Authors: I. Gusti Agung Aditya Surya Wibawa, Andri Syafriya, Beiruny Syam

Abstract:

Gumai Formation which acts as regional seal for Talang Akar Formation becomes one of the most prolific reservoir in South Sumatra Basin and the primary exploration target in this area. Marine conditions were eventually established during the continuation of transgression sequence leads an open marine facies deposition in Early Miocene. Marine clastic deposits where calcareous shales, claystone and siltstones interbedded with fine-grained calcareous and glauconitic sandstones are the domination of lithology which targeted as the hydrocarbon reservoir. All this time, the main objective of PetroChina’s exploration and production in Betara area is only from Lower Talang Akar Formation. Successful testing in some exploration wells which flowed gas & condensate from Gumai Formation, opened the opportunity to optimize new reservoir objective in Betara area. Limitation of conventional wireline logs data in Gumai interval is generating technical challenge in term of geological approach. A utilization of Gas While Drilling indicator initiated with the objective to determine the next Gumai reservoir candidate which capable to increase Jabung hydrocarbon discoveries. This paper describes how Gas While Drilling indicator is processed to generate potential and non-potential zone by cut-off analysis. Validation which performed by correlation and comparison with well logs, Drill Stem Test (DST), and Reservoir Performance Monitor (RPM) data succeed to observe Gumai reservoir in Betara Complex. After we integrated all of data, we are able to generate a Betara Complex potential map and overlaid with reservoir characterization distribution as a part of risk assessment in term of potential zone presence. Mud log utilization and geophysical data information successfully covered the geological challenges in this study.

Keywords: Gumai, gas while drilling, classification, reservoir, potential

Procedia PDF Downloads 355

Authors: Zunaira Zaman, David Bohannon

Abstract:

This study explores the application of machine learning (ML) for detecting security vulnerabilities in source code. The research aims to assist organizations with large application portfolios and limited security testing capabilities in prioritizing security activities. ML-based approaches offer benefits such as increased confidence scores, false positives and negatives tuning, and automated feedback. The initial approach using natural language processing techniques to extract features achieved 86% accuracy during the training phase but suffered from overfitting and performed poorly on unseen datasets during testing. To address these issues, the study proposes using the abstract syntax tree (AST) for Java and C++ codebases to capture code semantics and structure and generate path-context representations for each function. The Code2Vec model architecture is used to learn distributed representations of source code snippets for training a machine-learning classifier for vulnerability prediction. The study evaluates the performance of the proposed methodology using two datasets and compares the results with existing approaches. The Devign dataset yielded 60% accuracy in predicting vulnerable code snippets and helped resist overfitting, while the Juliet Test Suite predicted specific vulnerabilities such as OS-Command Injection, Cryptographic, and Cross-Site Scripting vulnerabilities. The Code2Vec model achieved 75% accuracy and a 98% recall rate in predicting OS-Command Injection vulnerabilities. The study concludes that even partial AST representations of source code can be useful for vulnerability prediction. The approach has the potential for automated intelligent analysis of source code, including vulnerability prediction on unseen source code. State-of-the-art models using natural language processing techniques and CNN models with ensemble modelling techniques did not generalize well on unseen data and faced overfitting issues. However, predicting vulnerabilities in source code using machine learning poses challenges such as high dimensionality and complexity of source code, imbalanced datasets, and identifying specific types of vulnerabilities. Future work will address these challenges and expand the scope of the research.

Keywords: code embeddings, neural networks, natural language processing, OS command injection, software security, code properties

Procedia PDF Downloads 107

Authors: Lidija R. Jevrić, Sanja O. Podunavac-Kuzmanović, Strahinja Z. Kovačević

Abstract:

Quantitative structure-retention relationship (QSRR) analysis was used to predict the chromatographic behavior of s-triazine derivatives by using theoretical descriptors computed from the chemical structure. Fundamental basis of the reported investigation is to relate molecular topological descriptors with chromatographic behavior of s-triazine derivatives obtained by reversed-phase (RP) thin layer chromatography (TLC) on silica gel impregnated with paraffin oil and applied ethanol-water (φ = 0.5-0.8; v/v). Retention parameter (RM0) of 14 investigated s-triazine derivatives was used as dependent variable while simple connectivity index different orders were used as independent variables. The best QSRR model for predicting RM0 value was obtained with simple third order connectivity index (3χ) in the second-degree polynomial equation. Numerical values of the correlation coefficient (r=0.915), Fisher's value (F=28.34) and root mean square error (RMSE = 0.36) indicate that model is statistically significant. In order to test the predictive power of the QSRR model leave-one-out cross-validation technique has been applied. The parameters of the internal cross-validation analysis (r2CV=0.79, r2adj=0.81, PRESS=1.89) reflect the high predictive ability of the generated model and it confirms that can be used to predict RM0 value. Multivariate classification technique, hierarchical cluster analysis (HCA), has been applied in order to group molecules according to their molecular connectivity indices. HCA is a descriptive statistical method and it is the most frequently used for important area of data processing such is classification. The HCA performed on simple molecular connectivity indices obtained from the 2D structure of investigated s-triazine compounds resulted in two main clusters in which compounds molecules were grouped according to the number of atoms in the molecule. This is in agreement with the fact that these descriptors were calculated on the basis of the number of atoms in the molecule of the investigated s-triazine derivatives.

Keywords: s-triazines, QSRR, chemometrics, chromatography, molecular descriptors

Procedia PDF Downloads 393

Authors: Pranita Shivankar, Arun Suryawanshi, Prabodhachandra Deshmukh, S. V. C. Kameswara Rao

Abstract:

Many undesirable significant changes in landscapes and the regions in the vicinity of historically important structures occur as impacts due to anthropogenic activities over a period of time. A better understanding of such influences using recently developed satellite remote sensing techniques helps in planning the strategies for minimizing the negative impacts on the existing environment. In 1982, a fossilized hominid skull cap was discovered at a site located along the northern bank of the east-west flowing river Narmada in the village Hathnora. Close to the same site, the presence of Late Acheulian and Middle Palaeolithic tools have been discovered in the immediately overlying pebbly gravel, suggesting that the ‘Narmada skull’ may be from the Middle Pleistocene age. The reviews of recently carried out research studies relevant to hominid remains all over the world from Late Acheulian and Middle Palaeolithic sites suggest succession and contemporaneity of cultures there, enhancing the importance of Hathnora as a rare precious site. In this context, the maximum likelihood classification using digital interpretation techniques was carried out for this study area using the satellite imagery from Landsat ETM+ for the year 2006 and Landsat TM (OLI and TIRS) for the year 2016. The overall accuracy of Land Use Land Cover (LULC) classification of 2016 imagery was around 77.27% based on ground truth data. The significant reduction in the main river course and agricultural activities and increase in the built-up area observed in remote sensing data analysis are undoubtedly the outcome of human encroachments in the vicinity of the eminent heritage site.

Keywords: cultural succession, digital interpretation, Hathnora, Homo Sapiens, Late Acheulian, Middle Palaeolithic

Procedia PDF Downloads 172

Authors: M. Baratieri, M. Beccali, S. Corradino, B. Di Pietra, C. La Grassa, F. Monteleone, G. Morosinotto, G. Puglisi

Abstract:

The energy renovation of existing buildings represents an important opportunity to increase the decarbonization and the sustainability of urban environments. In this context, the work carried out has the objective of demonstrating the technical and economic feasibility of an energy renovate of a public building destined for offices located on the island of Lampedusa in the Mediterranean Sea. By applying the Italian transpositions of European Directives 2010/31/EU and 2009/28/EC, the building has been renovated from the current energy requirements of 111.7 kWh/m² to 16.4 kWh/m². The result achieved classifies the building as nZEB (nearly Zero Energy Building) according to the Italian national definition. The analysis was carried out using in parallel a quasi-stationary software, normally used in the professional field, and a dynamic simulation model often used in the academic world. The proposed interventions cover the components of the building’s envelope, the heating-cooling system and the supply of energy from renewable sources. In these latter points, the analysis has focused more on assessing two aspects that affect the supply of renewable energy. The first concerns the use of advanced logic control systems for air conditioning units in order to increase photovoltaic self-consumption. With these adjustments, a considerable increase in photovoltaic self-consumption and a decrease in the electricity exported to the Island's electricity grid have been obtained. The second point concerned the evaluation of the building's energy classification considering the real efficiency of the heating-cooling plant. Normally the energy plants have lower operational efficiency than the designed one due to multiple reasons; the decrease in the energy classification of the building for this factor has been quantified. This study represents an important example for the evaluation of the best interventions for the energy renovation of buildings in the Mediterranean Climate and a good description of the correct methodology to evaluate the resulting improvements.

Keywords: heat pumps, HVAC systems, nZEB renovation, renewable energy sources

Procedia PDF Downloads 451

Authors: Massimo Miccoli, Luca Marangoni, Alberto Aniello Scaringi, Alessandro Marceddu, Alessandro Amicone

Abstract:

The susceptibility of deep neural networks (DNNs) to adversarial manipulations is a recognized challenge within the computer vision domain. Adversarial examples, crafted by adding subtle yet malicious alterations to benign images, exploit this vulnerability. Various defense strategies have been proposed to safeguard DNNs against such attacks, stemming from diverse research hypotheses. Building upon prior work, our approach involves the utilization of autoencoder models. Autoencoders, a type of neural network, are trained to learn representations of training data and reconstruct inputs from these representations, typically minimizing reconstruction errors like mean squared error (MSE). Our autoencoder was trained on a dataset of benign examples; learning features specific to them. Consequently, when presented with significantly perturbed adversarial examples, the autoencoder exhibited high reconstruction errors. The architecture of the autoencoder was tailored to the dimensions of the images under evaluation. We considered various image sizes, constructing models differently for 256x256 and 512x512 images. Moreover, the choice of the computer vision model is crucial, as most adversarial attacks are designed with specific AI structures in mind. To mitigate this, we proposed a method to replace image-specific dimensions with a structure independent of both dimensions and neural network models, thereby enhancing robustness. Our multi-modal autoencoder reconstructs the spectral representation of images across the red-green-blue (RGB) color channels. To validate our approach, we conducted experiments using diverse datasets and subjected them to adversarial attacks using models such as ResNet50 and ViT_L_16 from the torch vision library. The autoencoder extracted features used in a classification model, resulting in an MSE (RGB) of 0.014, a classification accuracy of 97.33%, and a precision of 99%.

Keywords: adversarial attacks, malicious images detector, binary classifier, multimodal transformer autoencoder

Procedia PDF Downloads 113

Authors: Sahand Golmohammadi, Sana Hosseini Shirazi

Abstract:

Because the status and mechanical parameters of discontinuities in the rock mass are included in the calculations, various methods of rock engineering classification are often used as a starting point for the design of different types of structures. The Q-system is one of the most frequently used methods for stability analysis and determination of support systems of underground structures in rock, including tunnel. In this method, six main parameters of the rock mass, namely, the rock quality designation (RQD), joint set number (Jn), joint roughness number (Jr), joint alteration number (Ja), joint water parameter (Jw) and stress reduction factor (SRF) are required. In this regard, in order to achieve a reasonable and optimal design, identifying the effective parameters for the stability of the mentioned structures is one of the most important goals and the most necessary actions in rock engineering. Therefore, it is necessary to study the relationships between the parameters of a system and how they interact with each other and, ultimately, the whole system. In this research, it has attempted to determine the most effective parameters (key parameters) from the six parameters of rock mass in the Q-system using the rock engineering system (RES) method to improve the relationships between the parameters in the calculation of the Q value. The RES system is, in fact, a method by which one can determine the degree of cause and effect of a system's parameters by making an interaction matrix. In this research, the geomechanical data collected from the water conveyor tunnel of Azad Dam were used to make the interaction matrix of the Q-system. For this purpose, instead of using the conventional methods that are always accompanied by defects such as uncertainty, the Q-system interaction matrix is coded using a technique that is actually a statistical analysis of the data and determining the correlation coefficient between them. So, the effect of each parameter on the system is evaluated with greater certainty. The results of this study show that the formed interaction matrix provides a reasonable estimate of the effective parameters in the Q-system. Among the six parameters of the Q-system, the SRF and Jr parameters have the maximum and minimum impact on the system, respectively, and also the RQD and Jw parameters have the maximum and minimum impact on the system, respectively. Therefore, by developing this method, we can obtain a more accurate relation to the rock mass classification by weighting the required parameters in the Q-system.

Keywords: Q-system, rock engineering system, statistical analysis, rock mass, tunnel

Procedia PDF Downloads 73

Authors: Vistorina Amputu, Katja Tielboerger, Nichola Knox

Abstract:

Primary productivity in dry savannahs is constraint by moisture availability and under increasing anthropogenic pressure. Thus, considering climate change and the unprecedented pace and scale of rangeland deterioration, methods for assessing the status of such rangelands should be easy to apply, yield reliable and repeatable results that can be applied over large spatial scales. Global and local scale monitoring of rangelands through satellite data and labor-intensive field measurements respectively, are limited in accurately assessing the spatiotemporal heterogeneity of vegetation dynamics to provide crucial information that detects degradation in its early stages. Fortunately, newly emerging techniques such as unmanned aerial vehicles (UAVs), associated miniaturized sensors and improving digital photogrammetric software provide an opportunity to transcend these limitations. Yet, they have not been extensively calibrated in natural systems to encompass their complexities if they are to be integrated for long-term monitoring. Limited research using drone technology has been conducted in arid savannas, for example to assess the health status of this dynamic two-layer vegetation ecosystem. In our study, we fill this gap by testing the relationship between UAV-estimated cover of rangeland functional attributes and field data collected in discrete sample plots in a Namibian dryland savannah along a degradation gradient. The first results are based on a supervised classification performed on the ultra-high resolution multispectral imagery to distinguish between rangeland functional attributes (bare, non-woody, and woody), with a relatively good match to the field observations. Integrating UAV-based observations to improve rangeland monitoring could greatly assist in climate-adapted rangeland management.

Keywords: arid savannah, degradation gradient, field observations, narrow-band sensor, supervised classification

Procedia PDF Downloads 134

Authors: Saeid Gharechelou, Ryutaro Tateishi

Abstract:

Earthquake is an inevitable catastrophic natural disaster. The damages of buildings and man-made structures, where most of the human activities occur are the major cause of casualties from earthquakes. A comparison of optical and SAR data is presented in the case of Kathmandu valley which was hardly shaken by 2015-Nepal Earthquake. Though many existing researchers have conducted optical data based estimated or suggested combined use of optical and SAR data for improved accuracy, however finding cloud-free optical images when urgently needed are not assured. Therefore, this research is specializd in developing SAR based technique with the target of rapid and accurate geospatial reporting. Should considers that limited time available in post-disaster situation offering quick computation exclusively based on two pairs of pre-seismic and co-seismic single look complex (SLC) images. The InSAR coherence pre-seismic, co-seismic and post-seismic was used to detect the change in damaged area. In addition, the ground truth data from field applied to optical data by random forest classification for detection of damaged area. The ground truth data collected in the field were used to assess the accuracy of supervised classification approach. Though a higher accuracy obtained from the optical data then integration by optical-SAR data. Limitation of cloud-free images when urgently needed for earthquak evevent are and is not assured, thus further research on improving the SAR based damage detection is suggested. Availability of very accurate damage information is expected for channelling the rescue and emergency operations. It is expected that the quick reporting of the post-disaster damage situation quantified by the rapid earthquake assessment should assist in channeling the rescue and emergency operations, and in informing the public about the scale of damage.

Keywords: Sentinel-1A data, Landsat-8, earthquake damage, InSAR, rapid damage monitoring, 2015-Nepal earthquake

Procedia PDF Downloads 172

Authors: Jeffrey Chase

Abstract:

Historically the profession of Occupational Therapy was closely tied to the treatment of those suffering from mental illness; more recently, and especially in the U.S., the percentage of OTs identifying as working in the mental health area has declined significantly despite the estimate that by 2020 behavioral health disorders will surpass physical illnesses as the major cause of disability worldwide. In the U.S. less than 10% of OTs identify themselves as working with the mentally ill and/or practicing in mental health settings. Such a decline has implications for both those suffering from mental illness and the profession of Occupational Therapy. One reason cited for the decline of OT in mental health has been the limited research in the discipline addressing mental health practice. Despite significant advances in technology and growth in the field of neuroscience, major institutions and funding sources such as the National Institute of Mental Health (NIMH) have noted that research into the etiology and treatment of mental illness have met with limited success over the past 25 years. One major reason posited by NIMH is that research has been limited by how we classify individuals, that being mostly on what is observable. A new classification system being developed by NIMH, the Research Domain Criteria (RDoc), has the goal to look beyond just descriptors of disorders for common neural, genetic, and physiological characteristics that cut across multiple supposedly separate disorders. The hope is that by classifying individuals along RDoC measures that both reliability and validity will improve resulting in greater advances in the field. As a result of this change NIH and NIMH will prioritize research funding to those projects using the RDoC model. Multiple disciplines across many different setting will be required for RDoC or similar classification systems to be developed. During this shift in research methodology OT has an opportunity to reassert itself into the research and treatment of mental illness, both in developing new ways to more validly classify individuals, and to document the legitimacy of previously ill-defined and validated disorders such as sensory integration.

Keywords: global mental health and neuroscience, research opportunities for ot, greater integration of ot in mental health research, research and funding opportunities, research domain criteria (rdoc)

Procedia PDF Downloads 275

Authors: Renuka G. Oka

Abstract:

From chowks to streets to verandahs to courtyards; residential open spaces are very significantly placed in traditional urban neighborhoods of India. At various levels of intersection, the open spaces with their attributes like juxtaposition with the built fabric, scale, climate sensitivity and response, multi-functionality, etc. reflect and respond to the patterns of human interactions. Also, these spaces tend to be quite well utilized. On the other hand, it is a common specter to see an imbalanced utilization of open spaces in newly/recently planned residential clusters. This is maybe due to lack of activity generators around or wrong locations or excess provisions or improper incorporation of aforementioned design attributes. These casual observations suggest the necessity for a systematic study of current residential open spaces. The exploratory study thus attempts to draw lessons through a structured inspection of residential open spaces to understand the effective environment as revealed through their use patterns. Here, residential open spaces are considered in a wider sense to incorporate all the un-built fabric around. These thus, include both use spaces and access space. For the study, open spaces in ten exemplary housing clusters/societies built during the last ten years across India are studied. A threefold inquiry is attempted in this direction. The first relates to identifying and determining the effects of various physical functions like space organization, size, hierarchy, thermal and optical comfort, etc. on the performance of residential open spaces. The second part sets out to understand socio-cultural variations in values, lifestyle, and beliefs which determine activity choices and behavioral preferences of users for respective residential open spaces. The third inquiry further observes the application of these research findings to the design process to derive meaningful and qualitative design advice. However, the study also emphasizes to develop a suitable framework of analysis and to carve out appropriate methods and approaches to probe into these aspects of the inquiry. Given this emphasis, a considerable portion of the research details out the conceptual framework for the study. This framework is supported by an in-depth search of available literature. The findings are worked out for design solutions which integrate the open space systems with the overall design process for residential clusters. The open spaces in residential areas present great complexities both in terms of their use patterns and determinants of their functional responses. The broad aim of the study is, therefore, to arrive at reconsideration of standards and qualitative parameters used by designers – on the basis of more substantial inquiry into the use patterns of open spaces in residential areas.

Keywords: open spaces, physical and social determinants, residential clusters, use patterns

Procedia PDF Downloads 148

Authors: O. J. Akinsola, Yinan Zheng, Rose Anorlu, F. T. Ogunsola, Lifang Hou, Robert Leo-Murphy

Abstract:

Cervical Cancer (CC) is the 2nd most common cancer among women living in low and middle-income countries, with no associated symptoms during formative periods. With the advancement and innovative medical research, there are numerous preventive measures being utilized, but the incidence of cervical cancer cannot be truncated with the application of only screening tests. The mortality associated with this invasive cervical cancer can be nipped in the bud through the important role of early-stage detection. This study research selected an array of different top features selection techniques which was aimed at developing a model that could validly diagnose the risk factors of cervical cancer. A retrospective clinic-based cohort study was conducted on 178 HIV-associated cervical cancer patients in Lagos University teaching Hospital, Nigeria (U54 data repository) in April 2022. The outcome measure was the automated prediction of the HIV-associated cervical cancer cases, while the predictor variables include: demographic information, reproductive history, birth control, sexual history, cervical cancer screening history for invasive cervical cancer. The proposed technique was assessed with R and Python programming software to produce the model by utilizing the classification algorithms for the detection and diagnosis of cervical cancer disease. Four machine learning classification algorithms used are: the machine learning model was split into training and testing dataset into ratio 80:20. The numerical features were also standardized while hyperparameter tuning was carried out on the machine learning to train and test the data. Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), and K-Nearest Neighbor (KNN). Some fitting features were selected for the detection and diagnosis of cervical cancer diseases from selected characteristics in the dataset using the contribution of various selection methods for the classification cervical cancer into healthy or diseased status. The mean age of patients was 49.7±12.1 years, mean age at pregnancy was 23.3±5.5 years, mean age at first sexual experience was 19.4±3.2 years, while the mean BMI was 27.1±5.6 kg/m2. A larger percentage of the patients are Married (62.9%), while most of them have at least two sexual partners (72.5%). Age of patients (OR=1.065, p<0.001**), marital status (OR=0.375, p=0.011**), number of pregnancy live-births (OR=1.317, p=0.007**), and use of birth control pills (OR=0.291, p=0.015**) were found to be significantly associated with HIV-associated cervical cancer. On top ten 10 features (variables) considered in the analysis, RF claims the overall model performance, which include: accuracy of (72.0%), the precision of (84.6%), a recall of (84.6%) and F1-score of (74.0%) while LR has: an accuracy of (74.0%), precision of (70.0%), recall of (70.0%) and F1-score of (70.0%). The RF model identified 10 features predictive of developing cervical cancer. The age of patients was considered as the most important risk factor, followed by the number of pregnancy livebirths, marital status, and use of birth control pills, The study shows that data mining techniques could be used to identify women living with HIV at high risk of developing cervical cancer in Nigeria and other sub-Saharan African countries.

Keywords: associated cervical cancer, data mining, random forest, logistic regression

Procedia PDF Downloads 83

Authors: Ahmed Elhussein, Hossam El-Zomor, Adel Alieldin, Mahmoud A. Afifi, Abdullah Elhusseiny, Hala Taha, Amal Refaat, Soha Ahmed, Mohamed S. Zagloul

Abstract:

Background: A majority of children with retinoblastoma (60%) have a disease in one eye only (unilateral disease). This is a retrospective study to evaluate two different treatment modalities in those patients for saving their lives and vision. Methods: Four hundred and four patients were diagnosed with unilateral intraocular retinoblastoma at Children’s Cancer, Hospital Egypt (CCHE) through the period of July/2007 until December/2017. Management strategies included primary enucleation versus ocular salvage treatment. Results: Patients presented with mean age 24.5 months with range (1.2-154.3 months). According to the international retinoblastoma classification, Group D (n=172, 42%) was the most common, followed by group E (n=142, 35%), group C (n=63, 16%), and group B (n=27, 7%). All patients were alive at the end of the study except four patients who died, with 5-years overall survival 98.3% [CI, (96.5-100%)]. Patients presented with advanced disease and poor visual prognosis (n=241, 59.6%) underwent primary enucleation with 6 cycles adjuvant chemotherapy if they had high-risk features in the enucleated eye; only four patients out of 241 ended-up either with extraocular metastasis (n=3) or death (n=1). While systemic chemotherapy and focal therapy were the primary treatment for those who presented with favorable disease status and good visual prognosis (n=163, 40.4%); seventy-seven patients of them (47%) ended up with a pre-defined event (enucleation, EBRT, off protocol chemotherapy or 2ry malignancy). Ocular survival for patients received primary chemotherapy + focal therapy was [50.9% (CI, 43.5-59.6%)] at 3 years and [46.9% (CI,39.3-56%)] at 5 years. Comparison between upfront enucleation and primary chemotherapy for occurrence of extraocular metastasis revealed that there was no statistical difference between them except in group D (p value). While for occurrence of death, no statistical difference in all classification groups. Conclusion: In retinoblastoma, primary chemotherapy is a reasonable option and has a good probability for ocular salvage without increasing the risk of metastasis in comparison to upfront enucleation except in group D.

Keywords: CCHE, chemotherapy, enucleation, retinoblastoma

Procedia PDF Downloads 155

Authors: Bianca W. O. Silva, Adriana C. M. Andrade, Gustavo C. M. Lucena, Virna M. S. Lima

Abstract:

Introduction: Basic life support (BLS) involves the immediate recognition of cardiopulmonary arrest. Each year, 359.400 and 275.000 individuals with cardiac arrest are attended in emergency departments in USA and Europe. Brazilian data shows that 200.000 cardiac arrests occur every year, and half of them out of the hospital. Medical schools around the world teach BLS in the first years of the course, but studies show that there is a decline of the knowledge as the years go by, affecting the chain of survival. The objective was to analyze the knowledge of medical students about BLS and the retention of this learning throughout the course. Methods: This study included 150 students who were at the second and third year of a medical school in Salvador, Bahia, Brazil. The instrument of data collection was a structured questionnaire composed of 20 questions based on the 2015 American Heart Association guideline. The Pearson Chi-square test was used in order to study the association between previous training, sex and semester with the degree of knowledge of the students. The Kruskal-Wallis test was used to evaluate the different yields obtained between the various semesters. The number of correct answers was described by average and quartiles. Results: Regarding the degree of knowledge, 19.6% of the female students reached the optimal classification, a better outcome than the achieved by the male participants. Of those with previous training, 33.33% were classified as good and optimal, none of the students reached the optimal classification and only 2.2% of them were classified as bad (those who did not have 52.6% of correct answers). The analysis of the degree of knowledge related to each semester revealed that the 5th semester had the highest outcome: 30.5%. However, the acquaintance presented by the semesters was generally unsatisfactory, since 50% of the students, or more, demonstrated knowledge levels classified as bad or regular. When confronting the different semesters and the achieved scores, the value of p was 0.831. Conclusion: It is important to focus on the training of medical professionals that are capable of facing emergency situations, improving the systematization of care, and thereby increasing the victims' possibility of survival.

Keywords: basic life support, cardiopulmonary ressucitacion, education, medical students

Procedia PDF Downloads 186

Authors: Oscar E. Cariceo, Claudia V. Casal

Abstract:

Machine learning offers a set of techniques to promote social work interventions and can lead to support decisions of practitioners in order to predict new behaviors based on data produced by the organizations, services agencies, users, clients or individuals. Machine learning techniques include a set of generalizable algorithms that are data-driven, which means that rules and solutions are derived by examining data, based on the patterns that are present within any data set. In other words, the goal of machine learning is teaching computers through 'examples', by training data to test specifics hypothesis and predict what would be a certain outcome, based on a current scenario and improve that experience. Machine learning can be classified into two general categories depending on the nature of the problem that this technique needs to tackle. First, supervised learning involves a dataset that is already known in terms of their output. Supervising learning problems are categorized, into regression problems, which involve a prediction from quantitative variables, using a continuous function; and classification problems, which seek predict results from discrete qualitative variables. For social work research, machine learning generates predictions as a key element to improving social interventions on complex social issues by providing better inference from data and establishing more precise estimated effects, for example in services that seek to improve their outcomes. This paper exposes the results of a classification algorithm to predict cyberbullying among adolescents. Data were retrieved from the National Polyvictimization Survey conducted by the government of Chile in 2017. A logistic regression model was created to predict if an adolescent would experience cyberbullying based on the interaction and behavior of gender, age, grade, type of school, and self-esteem sentiments. The model can predict with an accuracy of 59.8% if an adolescent will suffer cyberbullying. These results can help to promote programs to avoid cyberbullying at schools and improve evidence based practice.

Keywords: cyberbullying, evidence based practice, machine learning, social work research

Procedia PDF Downloads 168

Authors: M. P. Tripathi, Priti Tiwari

Abstract:

Soil and Water Assessment Tool (SWAT) is a distributed parameter continuous time model and was tested on daily and fortnightly basis for a small agricultural watershed (Dhangaon) of Chhattisgarh state in India. The SWAT model recently interfaced with ArcGIS and called as ArcSWAT. The watershed and sub-watershed boundaries, drainage networks, slope and texture maps were generated in the environment of ArcGIS of ArcSWAT. Supervised classification method was used for land use/cover classification from satellite imageries of the years 2009 and 2012. Manning's roughness coefficient 'n' for overland flow and channel flow and Fraction of Field Capacity (FFC) were calibrated for monsoon season of the years 2009 and 2010. The model was validated on a daily basis for the years 2011 and 2012 by using the observed daily rainfall and temperature data. Calibration and validation results revealed that the model was predicting the daily surface runoff and sediment yield satisfactorily. Sensitivity analysis showed that the annual sediment yield was inversely proportional to the overland and channel 'n' values whereas; annual runoff and sediment yields were directly proportional to the FFC. The model was also tested (calibrated and validated) for the fortnightly runoff and sediment yield for the year 2009-10 and 2011-12, respectively. Simulated values of fortnightly runoff and sediment yield for the calibration and validation years compared well with their observed counterparts. The calibration and validation results revealed that the ArcSWAT model could be used for identification of critical sub-watershed and for developing management scenarios for the Dhangaon watershed. Further, the model should be tested for simulating the surface runoff and sediment yield using generated rainfall and temperature before applying it for developing the management scenario for the critical or priority sub-watersheds.

Keywords: watershed, hydrologic and water quality, ArcSWAT model, remote sensing, GIS, runoff and sediment yield

Procedia PDF Downloads 379

Authors: Nathenal Thomas

Abstract:

In the food chain and in many other agricultural products, corn, also known as maize, which goes by the scientific name Zea mays subsp, is a widely produced agricultural product. Corn has the highest adaptability. It comes in many different types, is employed in many different industrial processes, and is more adaptable to different agro-climatic situations. In Ethiopia, maize is among the most widely grown crop. Small-scale corn farming may be a household's only source of food in developing nations like Ethiopia. The aforementioned data demonstrates that the country's requirement for this crop is excessively high, and conversely, the crop's productivity is very low for a variety of reasons. The most damaging disease that greatly contributes to this imbalance between the crop's supply and demand is the corn disease. The failure to diagnose diseases in maize plant until they are too late is one of the most important factors influencing crop output in Ethiopia. This study will aid in the early detection of such diseases and support farmers during the cultivation process, directly affecting the amount of maize produced. The diseases in maize plants, such as northern leaf blight and cercospora leaf spot, have distinct symptoms that are visible. This study aims to detect the most frequent and degrading maize diseases using the most efficiently used subset of machine learning technology, deep learning so, called Image Processing. Deep learning uses networks that can be trained from unlabeled data without supervision (unsupervised). It is a feature that simulates the exercises the human brain goes through when digesting data. Its applications include speech recognition, language translation, object classification, and decision-making. Convolutional Neural Network (CNN) for Image Processing, also known as convent, is a deep learning class that is widely used for image classification, image detection, face recognition, and other problems. it will also use this algorithm as the state-of-the-art for my research to detect maize diseases by photographing maize leaves using a mobile phone.

Keywords: CNN, zea mays subsp, leaf blight, cercospora leaf spot

Procedia PDF Downloads 74

Authors: Vitor M. Vilas Boas, Cleison D. Silva, Gustavo S. Mafra, Alexandre Trofino Neto

Abstract:

Motor imagery (MI) based brain-computer interfaces (BCI) uses event-related (de)synchronization (ERS/ ERD), typically recorded using electroencephalography (EEG), to translate brain electrical activity into control commands. To mitigate undesirable artifacts and noise measurements on EEG signals, methods based on band-pass filters defined by a specific frequency band (i.e., 8 – 30Hz), such as the Infinity Impulse Response (IIR) filters, are typically used. Spatial techniques, such as Common Spatial Patterns (CSP), are also used to estimate the variations of the filtered signal and extract features that define the imagined motion. The CSP effectiveness depends on the subject's discriminative frequency, and approaches based on the decomposition of the band of interest into sub-bands with smaller frequency ranges (SBCSP) have been suggested to EEG signals classification. However, despite providing good results, the SBCSP approach generally increases the computational cost of the filtering step in IM-based BCI systems. This paper proposes the use of the Fast Fourier Transform (FFT) algorithm in the IM-based BCI filtering stage that implements SBCSP. The goal is to apply the FFT algorithm to reduce the computational cost of the processing step of these systems and to make them more efficient without compromising classification accuracy. The proposal is based on the representation of EEG signals in a matrix of coefficients resulting from the frequency decomposition performed by the FFT, which is then submitted to the SBCSP process. The structure of the SBCSP contemplates dividing the band of interest, initially defined between 0 and 40Hz, into a set of 33 sub-bands spanning specific frequency bands which are processed in parallel each by a CSP filter and an LDA classifier. A Bayesian meta-classifier is then used to represent the LDA outputs of each sub-band as scores and organize them into a single vector, and then used as a training vector of an SVM global classifier. Initially, the public EEG data set IIa of the BCI Competition IV is used to validate the approach. The first contribution of the proposed method is that, in addition to being more compact, because it has a 68% smaller dimension than the original signal, the resulting FFT matrix maintains the signal information relevant to class discrimination. In addition, the results showed an average reduction of 31.6% in the computational cost in relation to the application of filtering methods based on IIR filters, suggesting FFT efficiency when applied in the filtering step. Finally, the frequency decomposition approach improves the overall system classification rate significantly compared to the commonly used filtering, going from 73.7% using IIR to 84.2% using FFT. The accuracy improvement above 10% and the computational cost reduction denote the potential of FFT in EEG signal filtering applied to the context of IM-based BCI implementing SBCSP. Tests with other data sets are currently being performed to reinforce such conclusions.

Keywords: brain-computer interfaces, fast Fourier transform algorithm, motor imagery, sub-band common spatial patterns

Procedia PDF Downloads 128

Authors: Tomiwa, Akinyemi Clement

Abstract:

Tropical regions experience diverse and complex precipitation patterns, posing significant challenges for accurate rainfall estimation and forecasting. This study addresses the problem of effectively classifying tropical rain types and refining the Z-R (Reflectivity-Rain Rate) relationship to enhance rainfall estimation accuracy. Through a combination of remote sensing, meteorological analysis, and machine learning, the research aims to develop an advanced classification framework capable of distinguishing between different types of tropical rain based on their unique characteristics. This involves utilizing high-resolution satellite imagery, radar data, and atmospheric parameters to categorize precipitation events into distinct classes, providing a comprehensive understanding of tropical rain systems. Additionally, the study seeks to improve the Z-R relationship, a crucial aspect of rainfall estimation. One year of rainfall data was analyzed using a Micro Rain Radar (MRR) located at The Federal University of Technology Akure, Nigeria, measuring rainfall parameters from ground level to a height of 4.8 km with a vertical resolution of 0.16 km. Rain rates were classified into low (stratiform) and high (convective) based on various microstructural attributes such as rain rates, liquid water content, Drop Size Distribution (DSD), average fall speed of the drops, and radar reflectivity. By integrating diverse datasets and employing advanced statistical techniques, the study aims to enhance the precision of Z-R models, offering a more reliable means of estimating rainfall rates from radar reflectivity data. This refined Z-R relationship holds significant potential for improving our understanding of tropical rain systems and enhancing forecasting accuracy in regions prone to heavy precipitation.

Keywords: remote sensing, precipitation, drop size distribution, micro rain radar

Procedia PDF Downloads 35

Authors: Fidelia A. Orji, Julita Vassileva

Abstract:

This research aims to develop machine learning models for students' academic performance and study strategy prediction, which could be generalized to all courses in higher education. Key learning attributes (intrinsic, extrinsic, autonomy, relatedness, competence, and self-esteem) used in building the models are chosen based on prior studies, which revealed that the attributes are essential in students’ learning process. Previous studies revealed the individual effects of each of these attributes on students’ learning progress. However, few studies have investigated the combined effect of the attributes in predicting student study strategy and academic performance to reduce the dropout rate. To bridge this gap, we used Scikit-learn in python to build five machine learning models (Decision Tree, K-Nearest Neighbour, Random Forest, Linear/Logistic Regression, and Support Vector Machine) for both regression and classification tasks to perform our analysis. The models were trained, evaluated, and tested for accuracy using 924 university dentistry students' data collected by Chilean authors through quantitative research design. A comparative analysis of the models revealed that the tree-based models such as the random forest (with prediction accuracy of 94.9%) and decision tree show the best results compared to the linear, support vector, and k-nearest neighbours. The models built in this research can be used in predicting student performance and study strategy so that appropriate interventions could be implemented to improve student learning progress. Thus, incorporating strategies that could improve diverse student learning attributes in the design of online educational systems may increase the likelihood of students continuing with their learning tasks as required. Moreover, the results show that the attributes could be modelled together and used to adapt/personalize the learning process.

Keywords: classification models, learning strategy, predictive modeling, regression models, student academic performance, student motivation, supervised machine learning

Procedia PDF Downloads 128

Authors: Semire Yekta

Abstract:

Many online retailers set up manual review teams to overcome the limitations of automated online fraud detection systems. This study critically examines the strategies they adapt in their decision-making process to set apart fraudulent individuals from non-fraudulent online shoppers. The study uses a mix method research approach. 32 in-depth interviews have been conducted alongside with participant observation and auto-ethnography. The study found out that all steps of the decision-making process are significantly affected by a level of subjectivity, personal understandings of online fraud, preferences and judgments and not necessarily by objectively identifiable facts. Rather clearly knowing who the fraudulent individuals are, the team members have to predict whether they think the customer might be a fraudster. Common strategies used are relying on the classification and fraud scorings in the automated fraud detection systems, weighing up arguments for and against the customer and making a decision, using cancellation to test customers’ reaction and making use of personal experiences and “the sixth sense”. The interaction in the team also plays a significant role given that some decisions turn into a group discussion. While customer data represent the basis for the decision-making, fraud management teams frequently make use of Google search and Google Maps to find out additional information about the customer and verify whether the customer is the person they claim to be. While this, on the one hand, raises ethical concerns, on the other hand, Google Street View on the address and area of the customer puts customers living in less privileged housing and areas at a higher risk of being classified as fraudsters. Phone validation is used as a final measurement to make decisions for or against the customer when previous strategies and Google Search do not suffice. However, phone validation is also characterized by individuals’ subjectivity, personal views and judgment on customer’s reaction on the phone that results in a final classification as genuine or fraudulent.

Keywords: online fraud, data mining, manual review, social construction

Procedia PDF Downloads 343

Authors: Hala M. El-hanbuli, Mohammed F. Darweesh

Abstract:

The incidence rates of any tumor vary hugely with geographical location. Basal Cell Carcinoma (BCC) is one of the most common skin cancer that has many histopathologic subtypes. Objective: The aim was to study the histopathological features of BCC cases that were received in the Pathology Department, Kasr El-Aini hospital, Cairo University, Egypt during the period from Jan 2004 to Dec 2013 and to evaluate the clinical characters through the patient data available in the request sheets. Methods: Slides and data of BCC cases were collected from the archives of the pathology department, Kasr El-Aini hospital. Revision of all available slides and histological classification of BCC according to WHO (2006) was done. Results: A total number of 310 cases of BCC representing about 65% from the total number of malignant skin tumors examined during the 10-years duration in the department. The age ranged from 8 to 84 years, the mean age was (55.7 ± 15.5). Most of the patients (85%) were above the age of 40 years. There was a slight male predominance (55%). Ulcerated BCC was the most common gross picture (60%), followed by nodular lesion (30%) and finally the ulcerated nodule (10%). Most of the lesions situated in the high-risk sites (77%) where the nose was the most common site (35%) followed by the periocular area (22%), then periauricular (15%) and finally perioral (5%). No lesion was reported outside the head. The tumor size was less than 2 centimeters in 65% of cases, and from 2-5 centimeters in the lesions' greatest dimension in the rest of cases. Histopathological reclassification revealed that the nodular BCC was the most common (68%) followed by the pigmented nodular (18.75%). The histologic high-risk groups represented (7.5%) about half of them (3.75%) being basosquamous carcinoma. The total incidence for multiple BCC and 2nd primary was 12%. Recurrent BCC represented 8%. All of the recurrent lesions of BCC belonged to the histologic high-risk group. Conclusion: Basal Cell Carcinoma is the most common skin cancer in the 10-year survey. Histopathological diagnosis and classification of BCC cases are essential for the determination of the tumor type and its biological behavior.

Keywords: basal cell carcinoma, high risk, histopathological features, statistical analysis

Procedia PDF Downloads 149

Authors: Kisik Song, Sungjoo Lee

Abstract:

With the growing importance of intangible assets recently, the impact of patent infringement on the business of a company has become more evident. Accordingly, it is essential for firms to estimate the risk of patent infringement risk before developing a technology and create new technology ideas to avoid the risk. Recognizing the needs, several attempts have been made to help develop new technology opportunities and most of them have focused on identifying emerging vacant technologies from patent analysis. In these studies, the IPC (International Patent Classification) system or keywords from text-mining application to patent documents was generally used to define vacant technologies. Unlike those studies, this study adopted F-term, which classifies patent documents according to the technical features of the inventions described in them. Since the technical features are analyzed by various perspectives by F-term, F-term provides more detailed information about technologies compared to IPC while more systematic information compared to keywords. Therefore, if well utilized, it can be a useful guideline to create a new technology idea. Recognizing the potential of F-term, this paper aims to suggest a novel approach to developing new technology ideas to avoid patent infringement based on F-term. For this purpose, we firstly collected data about F-term and then applied text-mining to the descriptions about classification criteria and attributes. From the text-mining results, we could identify other technologies with similar technical features of the existing one, the patented technology. Finally, we compare the technologies and extract the technical features that are commonly used in other technologies but have not been used in the existing one. These features are presented in terms of “purpose”, “function”, “structure”, “material”, “method”, “processing and operation procedure” and “control means” and so are useful for creating new technology ideas that help avoid infringing patent rights of other companies. Theoretically, this is one of the earliest attempts to adopt F-term to patent analysis; the proposed methodology can show how to best take advantage of F-term with the wealth of technical information. In practice, the proposed methodology can be valuable in the ideation process for successful product and service innovation without infringing the patents of other companies.

Keywords: patent infringement, new technology ideas, patent analysis, F-term

Procedia PDF Downloads 269

Authors: George Zhou, Yunchan Chen, Candace Chien

Abstract:

Kidney replacement therapy is the current standard of care for end-stage renal diseases. In-center or home hemodialysis remains an integral component of the therapeutic regimen. Arteriovenous fistulas (AVF) make up the vascular circuit through which blood is filtered and returned. Naturally, AVF patency determines whether adequate clearance and filtration can be achieved and directly influences clinical outcomes. Our aim was to build a deep learning model for automated AVF stenosis screening based on the sound of blood flow through the AVF. A total of 311 patients with AVF were enrolled in this study. Blood flow sounds were collected using a digital stethoscope. For each patient, blood flow sounds were collected at 6 different locations along the patient’s AVF. The 6 locations are artery, anastomosis, distal vein, middle vein, proximal vein, and venous arch. A total of 1866 sounds were collected. The blood flow sounds are labeled as “patent” (normal) or “stenotic” (abnormal). The labels are validated from concurrent ultrasound. Our dataset included 1527 “patent” and 339 “stenotic” sounds. We show that blood flow sounds vary significantly along the AVF. For example, the blood flow sound is loudest at the anastomosis site and softest at the cephalic arch. Contextualizing the sound with location metadata significantly improves classification performance. How to encode and incorporate categorical metadata is an active area of research1. Herein, we study ordinal (i.e., integer) encoding schemes. The numerical representation is concatenated to the flattened feature vector. We train a vision transformer (ViT) on spectrogram image representations of the sound and demonstrate that using scalar multiples of our integer encodings improves classification performance. Models are evaluated using a 10-fold cross-validation procedure. The baseline performance of our ViT without any location metadata achieves an AuROC and AuPRC of 0.68 ± 0.05 and 0.28 ± 0.09, respectively. Using the following encodings of Artery:0; Arch: 1; Proximal: 2; Middle: 3; Distal 4: Anastomosis: 5, the ViT achieves an AuROC and AuPRC of 0.69 ± 0.06 and 0.30 ± 0.10, respectively. Using the following encodings of Artery:0; Arch: 10; Proximal: 20; Middle: 30; Distal 40: Anastomosis: 50, the ViT achieves an AuROC and AuPRC of 0.74 ± 0.06 and 0.38 ± 0.10, respectively. Using the following encodings of Artery:0; Arch: 100; Proximal: 200; Middle: 300; Distal 400: Anastomosis: 500, the ViT achieves an AuROC and AuPRC of 0.78 ± 0.06 and 0.43 ± 0.11. respectively. Interestingly, we see that using increasing scalar multiples of our integer encoding scheme (i.e., encoding “venous arch” as 1,10,100) results in progressively improved performance. In theory, the integer values do not matter since we are optimizing the same loss function; the model can learn to increase or decrease the weights associated with location encodings and converge on the same solution. However, in the setting of limited data and computation resources, increasing the importance at initialization either leads to faster convergence or helps the model escape a local minimum.

Keywords: arteriovenous fistula, blood flow sounds, metadata encoding, deep learning

Procedia PDF Downloads 88

Authors: Naina Mahajan, Bikram Pal Kaur

Abstract:

The purpose of traffic accident analysis is to find the possible causes of an accident. Road accidents cannot be totally prevented but by suitable traffic engineering and management the accident rate can be reduced to a certain extent. This paper discusses the classification techniques C4.5 and ID3 using the WEKA Data mining tool. These techniques use on the NH (National highway) dataset. With the C4.5 and ID3 technique it gives best results and high accuracy with less computation time and error rate.

Keywords: C4.5, ID3, NH(National highway), WEKA data mining tool

Procedia PDF Downloads 338