Search results for: feature clustering

Authors: Hadeer R. M. Tawfik, Rania A. K. Birry, Amani A. Saad

Abstract:

Eyes are considered to be the most sensitive and important organ for human being. Thus, any eye disorder will affect the patient in all aspects of life. Cataract is one of those eye disorders that lead to blindness if not treated correctly and quickly. This paper demonstrates a model for automatic detection, classification, and grading of cataracts based on image processing techniques and artificial intelligence. The proposed system is developed to ease the cataract diagnosis process for both ophthalmologists and patients. The wavelet transform combined with 2D Log Gabor Wavelet transform was used as feature extraction techniques for a dataset of 120 eye images followed by a classification process that classified the image set into three classes; normal, early, and advanced stage. A comparison between the two used classifiers, the support vector machine SVM and the artificial neural network ANN were done for the same dataset of 120 eye images. It was concluded that SVM gave better results than ANN. SVM success rate result was 96.8% accuracy where ANN success rate result was 92.3% accuracy.

Keywords: cataract, classification, detection, feature extraction, grading, log-gabor, neural networks, support vector machines, wavelet

Procedia PDF Downloads 308

Authors: M. K. Witek-Hajduk, T. M. Napiórkowski

Abstract:

A number of studies discussed the topic of benefits of retailers-manufacturers cooperation and coopetition. However, there are only few publications focused on the benefits of cooperation and coopetition between retailers and their suppliers of durable consumer goods; especially in the context of business model of cooperating partners. This paper aims to provide a clustering approach to segment retailers selling consumer durables according to the benefits they obtain from their cooperation with key manufacturers and differentiate the said retailers’ in term of the business models of cooperating partners. For the purpose of the study, a survey (with a CATI method) collected data on 603 consumer durables retailers present on the Polish market. Retailers are clustered both, with hierarchical and non-hierarchical methods. Five distinctive groups of consumer durables’ retailers are (based on the studied benefits) identified using the two-stage clustering approach. The clusters are then characterized with a set of exogenous variables, key of which are business models employed by the retailer and its partnering key manufacturer. The paper finds that the a combination of a medium sized retailer classified as an Integrator with a chiefly domestic capital and a manufacturer categorized as a Market Player will yield the highest benefits. On the other side of the spectrum is medium sized Distributor retailer with solely domestic capital – in this case, the business model of the cooperating manufactrer appears to be irreleveant. This paper is the one of the first empirical study using cluster analysis on primary data that defines the types of cooperation between consumer durables’ retailers and manufacturers – their key suppliers. The analysis integrates a perspective of both retailers’ and manufacturers’ business models and matches them with individual and joint benefits.

Keywords: benefits of cooperation, business model, cluster analysis, retailer-manufacturer cooperation

Procedia PDF Downloads 239

Authors: Migena Ceyhan, Zeynep Orhan, Dimitrios Karras

Abstract:

People in modern societies are continuously sharing their experiences, emotions, and thoughts in different areas of life. The information reaches almost everyone in real-time and can have an important impact in shaping people’s way of living. This phenomenon is very well recognized and advantageously used by the market representatives, trying to earn the most from this means. Given the abundance of information, people and organizations are looking for efficient tools that filter the countless data into important information, ready to analyze. This paper is a modest contribution in this field, describing the process of automatically classifying social media comments in the Turkish language into positive or negative. Once data is gathered and preprocessed, feature sets of selected single words or groups of words are build according to the characteristics of language used in the texts. These features are used later to train, and test a system according to different machine learning algorithms (Naïve Bayes, Sequential Minimal Optimization, J48, and Bayesian Linear Regression). The resultant high accuracies can be important feedback for decision-makers to improve the business strategies accordingly.

Keywords: feature selection, machine learning, natural language processing, sentiment analysis, social media reviews

Procedia PDF Downloads 131

Authors: Marwan Jarrah

Abstract:

This research paper provides two pieces of phonological and syntactic evidence from Arabic for biolinguistics perspective of language processing. The first piece of evidence concerns the instances where a singular noun is converted to a plural noun in Arabic. Based on the findings of several research papers, this study shows that a singular word does not lose any of its moras when it is pluralized either regularly or irregularly. This mora conservation principle complies with the general physical law of the conservation of mass which states that mass is neither created nor destroyed but changed from one form into another. The second piece of evidence concerns the observation that when the object in some Arabic dialects including Jordanian Arabic and Najdi Arabic is a topic and positioned in situ (i.e. after the verb), the verb agrees with it, something that generates an agreeing inflection marker of the verb that agrees in Number, Person, and Gender with the in-situ topicalized object. This interaction between the verb and the object in such cases is invoked because of the extra feature the object bears, i.e. TOPIC feature. We suggest that such an interaction complies with the general natural law that elements become active when they, e.g., get an additional electron, when the mass number is not equal to the atomic number.

Keywords: biolinguistics, Arabic, physics, interaction

Procedia PDF Downloads 212

Authors: Niddal Imam, Vassilios G. Vassilakis

Abstract:

After analysing messages posted by health-related spam campaigns in Twitter Arabic hashtags, we found that these campaigns use unique hijacked accounts (we call them adversarial hijacked accounts) as adversarial examples to fool deployed ML-based spam detectors. Existing ML-based models build a behaviour profile for each user to detect hijacked accounts. This approach is not applicable for detecting spam in Twitter hashtags since they are computationally expensive. Hence, we propose an adversary-aware ML-based detector, which includes a newly designed feature (avg posts) to improve the detection of spam tweets posted by the adversarial hijacked accounts at a tweet-level in trending hashtags. The proposed detector was designed considering three key points: robustness, adaptability, and interpretability. The new feature leverages the account’s temporal patterns (i.e., account age and number of posts). It is faster to compute compared to features discussed in the literature and improves the accuracy of detecting the identified hijacked accounts by 73%.

Keywords: Twitter spam detection, adversarial examples, evasion attack, adversarial concept drift, account hijacking, trending hashtag

Procedia PDF Downloads 55

Authors: Bing-Wen Chen, Alvin W. Y. Su, Yu-Lin Wang

Abstract:

Electromyography (EMG) is one of the important indicators during exercise, as it is closely related to the level of muscle activations. This work quantifies the muscle conditions of the lower limbs in a constant workload exercise. Surface EMG signals of the vastus laterals (VL), vastus medialis (VM), rectus femoris (RF), gastrocnemius medianus (GM), gastrocnemius lateral (GL) and Soleus (SOL) were recorded from fourteen healthy males. The EMG signals were segmented in two phases: activation segment (AS) and relaxation segment (RS). Period entropy (PE), peak count (PC), zero crossing (ZC), wave length (WL), mean power frequency (MPF), median frequency (MDF) and root mean square (RMS) are calculated to provide the quantitative information of the measured EMG segments. The outcomes reveal that the PE, PC, ZC and RMS have significantly changed (p<.001); WL presents moderately changed (p<.01); MPF and MDF show no changed (p>.05) during exercise. The results also suggest that the RS is also preferred for performance evaluation, while the results of the extracted features in AS are usually affected directly by the amplitudes. It is further found that the VL exhibits the most significant changes within six muscles during pedaling exercise. The proposed work could be applied to quantify the stamina analysis and to predict the instant muscle status in athletes.

Keywords: electromyographic feature extraction, muscle status, pedaling exercise, relaxation segment

Procedia PDF Downloads 284

Authors: Edward Schwalb, Matthias Dehmer, Michael Schlenkrich, Farzaneh Taslimi, Ketron Mitchell-Wynne, Horen Kuecuekyan

Abstract:

We present a generalized feature extraction approach for supporting Machine Learning (ML) algorithms which perform tasks similar to Finite-Element Analysis (FEA). We report results for estimating the Head Injury Categorization (HIC) of vehicle engine compartments across various impact scenarios. Our experiments demonstrate that models learned using features derived with a simple discretization approach provide a reasonable approximation of a full simulation. We observe that Decision Trees could be as effective as Neural Networks for the HIC task. The simplicity and performance of the learned Decision Trees could offer a trade-off of a multiple order of magnitude increase in speed and cost improvement over full simulation for a reasonable approximation. When used as a complement to full simulation, the approach enables rapid approximate feedback to engineering teams before submission for full analysis. The approach produces mesh independent features and is further agnostic of the assembly structure.

Keywords: mechanical design validation, FEA, supervised decision tree, convolutional neural network.

Procedia PDF Downloads 119

Authors: Chusak Thanawattano, Roongroj Bhidayasiri

Abstract:

This paper proposes a novel feature set utilized for classifying the Parkinson tremor and essential tremor. Ten ET and ten PD subjects are asked to perform kinetic, postural and resting tests. The empirical mode decomposition (EMD) is used to decompose collected tremor signal to a set of intrinsic mode functions (IMF). The IMFs are used for reconstructing representative signals. The feature set is composed of peak frequencies of IMFs and reconstructed signals. Hypothesize that the dominant frequency components of subjects with PD and ET change in different directions for different tests, difference of peak frequencies of IMFs and reconstructed signals of pairwise based tests (kinetic-resting, kinetic-postural and postural-resting) are considered as potential features. Sets of features are used to train and test by classifier including the quadratic discriminant classifier (QLC) and the support vector machine (SVM). The best accuracy, the best sensitivity and the best specificity are 90%, 87.5%, and 92.86%, respectively.

Keywords: tremor, Parkinson, essential tremor, empirical mode decomposition, quadratic discriminant, support vector machine, peak frequency, auto-regressive, spectrum estimation

Procedia PDF Downloads 422

Authors: Rojin Bana Derakhshan, Abbas Toloie

Abstract:

For the purpose of optimizing and developing energy efficiency in building, it is required to recognize key elements of success in optimization of energy consumption before performing any actions. Surveying Principal Components is one of the most valuable result of Linear Algebra because the simple and non-parametric methods are become confusing. So that energy management system implemented according to energy management system international standard ISO50001:2011 and all energy parameters in building to be measured through performing energy auditing. In this essay by simulating used of data mining, the key impressive elements on energy saving in buildings to be determined. This approach is based on data mining statistical techniques using feature selection method and fuzzy logic and convert data from massive to compressed type and used to increase the selected feature. On the other side, influence portion and amount of each energy consumption elements in energy dissipation in percent are recognized as separated norm while using obtained results from energy auditing and after measurement of all energy consuming parameters and identified variables. Accordingly, energy saving solution divided into 3 categories, low, medium and high expense solutions.

Keywords: energy saving, key elements of success, optimization of energy consumption, data mining

Procedia PDF Downloads 446

Authors: Francisco Miroslav Ulloa-Stanojlovic, Gina Polo, Ricardo Augusto Dias

Abstract:

In Brazil, there is a high incidence of rabies in herbivores of economic interest (HEI) transmitted by the common vampire bat Desmodus rotundus, the presence of human rabies cases and the huge economic losses in the world's largest cattle industry, it is important to assist the National Program for Control of Rabies in herbivores in Brazil, that aims to reduce the incidence of rabies in HEI populations, mainly through epidemiological surveillance, vaccination of herbivores and control of vampire-bat roosts. Material and Methods: A spatiotemporal retrospective Kulldorff's spatial scan statistic based on a Poisson model and Monte Carlo simulation and an Anselin's Local Moran's I statistic were used to uncover spatial clustering of HEI rabies from 2000 – 2014. Results: Were identify three important clusters with significant year-to-year variation (Figure 1). In 2000, was identified one area of clustering in the North region, specifically in the State of Tocantins. Between the year 2000 and 2004, a cluster centered in the Midwest and Southeast region including the States of Goiás, Minas Gerais, Rio de Janeiro, Espirito Santo and São Paulo was prominent. And finally between 2000 and 2005 was found an important cluster in the North, Midwest and South region. Conclusions: The HEI rabies is endemic in the country, in addition, appears to be significant differences among the States according to their surveillance services, that may be difficulting the control of the disease, also other factors could be influencing in the maintenance of this problem like the lack of information of vampire-bat roosts identification, and limited human resources for realization of field monitoring. A review of the program control by the authorities it’s necessary.

Keywords: Brazil, Desmodus rotundus, herbivores, rabies

Procedia PDF Downloads 394

Authors: Zuhal Kurt, Meral Kurt, Bilge T. Bal, Kemal Ozkan

Abstract:

Shade matching and reproduction is the most important element of success in prosthetic dentistry. Until recently, shade matching procedure was implemented by dentists visual perception with the help of shade guides. Since many factors influence visual perception; tooth shade matching using visual devices (shade guides) is highly subjective and inconsistent. Subjective nature of this process has lead to the development of instrumental devices. Nowadays, colorimeters, spectrophotometers, spectroradiometers and digital image analysing systems are used for instrumental shade selection. Instrumental devices have advantages that readings are quantifiable, can obtain more rapidly and simply, objectively and precisely. However, these devices have noticeable drawbacks. For example, translucent structure and irregular surfaces of teeth lead to defects on measurement with these devices. Also between the results acquired by devices with different measurement principles may make inconsistencies. So, its obligatory to search for new methods for dental shade matching process. A computer-aided system device; digital camera has developed rapidly upon today. Currently, advances in image processing and computing have resulted in the extensive use of digital cameras for color imaging. This procedure has a much cheaper process than the use of traditional contact-type color measurement devices. Digital cameras can be taken by the place of contact-type instruments for shade selection and overcome their disadvantages. Images taken from teeth show morphology and color texture of teeth. In last decades, a new method was recommended to compare the color of shade tabs taken by a digital camera using color features. This method showed that visual and computer-aided shade matching systems should be used as concatenated. Recently using methods of feature extraction techniques are based on shape description and not used color information. However, color is mostly experienced as an essential property in depicting and extracting features from objects in the world around us. When local feature descriptors with color information are extended by concatenating color descriptor with the shape descriptor, that descriptor will be effective on visual object recognition and classification task. Therefore, the color descriptor is to be used in combination with a shape descriptor it does not need to contain any spatial information, which leads us to use local histograms. This local color histogram method is remain reliable under variation of photometric changes, geometrical changes and variation of image quality. So, coloring local feature extraction methods are used to extract features, and also the Scale Invariant Feature Transform (SIFT) descriptor used to for shape description in the proposed method. After the combination of these descriptors, the state-of-art descriptor named by Color-SIFT will be used in this study. Finally, the image feature vectors obtained from quantization algorithm are fed to classifiers such as Nearest Neighbor (KNN), Naive Bayes or Support Vector Machines (SVM) to determine label(s) of the visual object category or matching. In this study, SVM are used as classifiers for color determination and shade matching. Finally, experimental results of this method will be compared with other recent studies. It is concluded from the study that the proposed method is remarkable development on computer aided tooth shade determination system.

Keywords: classifiers, color determination, computer-aided system, tooth shade matching, feature extraction

Procedia PDF Downloads 410

Authors: Huayuan Lu, Chunfang Yang, Ma Zhu, Baojun Qi, Yaqiong Qiao, Jiangqian Xu

Abstract:

Image geolocalization has great application prospects in fields such as autonomous driving and virtual/augmented reality. In practical application scenarios, the size of the image to be located is not fixed; it is impractical to train different networks for all possible sizes. When its size does not match the size of the input of the descriptor extraction model, existing image geolocalization methods usually directly scale or crop the image in some common ways. This will result in the loss of some information important to the geolocalization task, thus affecting the performance of the image geolocalization method. For example, excessive down-sampling can lead to blurred building contour, and inappropriate cropping can lead to the loss of key semantic elements, resulting in incorrect geolocation results. To address this problem, this paper designs a learnable image resizer and proposes an arbitrary-sized image geolocation method. (1) The designed learnable image resizer employs the self-attention mechanism to enhance the geometric features of the resized image. Firstly, it applies bilinear interpolation to the input image and its feature maps to obtain the initial resized image and the resized feature maps. Then, SKNet (selective kernel net) is used to approximate the best receptive field, thus keeping the geometric shapes as the original image. And SENet (squeeze and extraction net) is used to automatically select the feature maps with strong contour information, enhancing the geometric features. Finally, the enhanced geometric features are fused with the initial resized image, to obtain the final resized images. (2) The proposed image geolocalization method embeds the above image resizer as a fronting layer of the descriptor extraction network. It not only enables the network to be compatible with arbitrary-sized input images but also enhances the geometric features that are crucial to the image geolocalization task. Moreover, the triplet attention mechanism is added after the first convolutional layer of the backbone network to optimize the utilization of geometric elements extracted by the first convolutional layer. Finally, the local features extracted by the backbone network are aggregated to form image descriptors for image geolocalization. The proposed method was evaluated on several mainstream datasets, such as Pittsburgh30K, Tokyo24/7, and Places365. The results show that the proposed method has excellent size compatibility and compares favorably to recently mainstream geolocalization methods.

Keywords: image geolocalization, self-attention mechanism, image resizer, geometric feature

Procedia PDF Downloads 195

Authors: Anshika Kankane, Dongshik Kang

Abstract:

Marine debris has a significant influence on coastal environments, damaging biodiversity, and causing loss and damage to marine and ocean sector. A functional cost-effective and automatic approach has been used to look up at this problem. Computer vision combined with a deep learning-based model is being proposed to identify and categorize marine debris of seven kinds on different beach locations of Japan. This research compares state-of-the-art deep learning models with a suggested model architecture that is utilized as a feature extractor for debris categorization. The model is being proposed to detect seven categories of litter using a manually constructed debris dataset, with the help of Mask R-CNN for instance segmentation and a shape matching network called HOGShape, which can then be cleaned on time by clean-up organizations using warning notifications of the system. The manually constructed dataset for this system is created by annotating the images taken by fixed KaKaXi camera using CVAT annotation tool with seven kinds of category labels. A pre-trained HOG feature extractor on LIBSVM is being used along with multiple templates matching on HOG maps of images and HOG maps of templates to improve the predicted masked images obtained via Mask R-CNN training. This system intends to timely alert the cleanup organizations with the warning notifications using live recorded beach debris data. The suggested network results in the improvement of misclassified debris masks of debris objects with different illuminations, shapes, viewpoints and litter with occlusions which have vague visibility.

Keywords: computer vision, debris, deep learning, fixed live camera images, histogram of gradients feature extractor, instance segmentation, manually annotated dataset, multiple template matching

Procedia PDF Downloads 80

Authors: Ana Neves, John Leander, Ignacio Gonzalez, Raid Karoumi

Abstract:

Systematic monitoring and inspection are needed to assess the present state of a structure and predict its future condition. If an irregularity is noticed, repair actions may take place and the adequate intervention will most probably reduce the future costs with maintenance, minimize downtime and increase safety by avoiding the failure of the structure as a whole or of one of its structural parts. For this to be possible decisions must be made at the right time, which implies using systems that can detect abnormalities in their early stage. In this sense, Structural Health Monitoring (SHM) is seen as an effective tool for improving the safety and reliability of infrastructures. This paper explores the decision-making problem in SHM regarding the maintenance of civil engineering structures. The aim is to assess the present condition of a bridge based exclusively on measurements using the suggested method in this paper, such that action is taken coherently with the information made available by the monitoring system. Artificial Neural Networks are trained and their ability to predict structural behavior is evaluated in the light of a case study where acceleration measurements are acquired from a bridge located in Stockholm, Sweden. This relatively old bridge is presently still in operation despite experiencing obvious problems already reported in previous inspections. The prediction errors provide a measure of the accuracy of the algorithm and are subjected to further investigation, which comprises concepts like clustering analysis and statistical hypothesis testing. These enable to interpret the obtained prediction errors, draw conclusions about the state of the structure and thus support decision making regarding its maintenance.

Keywords: artificial neural networks, clustering analysis, model-free damage detection, statistical hypothesis testing, structural health monitoring

Procedia PDF Downloads 188

Authors: P. V. Pramila , V. Mahesh

Abstract:

Pulmonary Function Tests are important non-invasive diagnostic tests to assess respiratory impairments and provides quantifiable measures of lung function. Spirometry is the most frequently used measure of lung function and plays an essential role in the diagnosis and management of pulmonary diseases. However, the test requires considerable patient effort and cooperation, markedly related to the age of patients esulting in incomplete data sets. This paper presents, a nonlinear model built using Multivariate adaptive regression splines and Random forest regression model to predict the missing spirometric features. Random forest based feature selection is used to enhance both the generalization capability and the model interpretability. In the present study, flow-volume data are recorded for N= 198 subjects. The ranked order of feature importance index calculated by the random forests model shows that the spirometric features FVC, FEF 25, PEF,FEF 25-75, FEF50, and the demographic parameter height are the important descriptors. A comparison of performance assessment of both models prove that, the prediction ability of MARS with the `top two ranked features namely the FVC and FEF 25 is higher, yielding a model fit of R2= 0.96 and R2= 0.99 for normal and abnormal subjects. The Root Mean Square Error analysis of the RF model and the MARS model also shows that the latter is capable of predicting the missing values of FEV1 with a notably lower error value of 0.0191 (normal subjects) and 0.0106 (abnormal subjects). It is concluded that combining feature selection with a prediction model provides a minimum subset of predominant features to train the model, yielding better prediction performance. This analysis can assist clinicians with a intelligence support system in the medical diagnosis and improvement of clinical care.

Keywords: FEV, multivariate adaptive regression splines pulmonary function test, random forest

Procedia PDF Downloads 288

Authors: A. Yamaoka, S. Yukawa

Abstract:

Mind wandering is a mental phenomenon of drifting away from a current task or external environment toward inner thought. This research examines the feature of mind wandering which people who have high creativity and high mental health engage in because it is expected that mind wandering which such kind of people engage in may not induce negative affect, although it can improve creativity. Sixty-seven participants were required to complete questionnaires which measured their creativity and mental health. After that, researchers conducted experience sampling method and measured the details of their mind wandering and the situation when mind wandering was generated in daily life for three days. The result showed that high creative people and high mental health people more think about positive things during mind wandering and less think about negative things. In further research, researchers will examine how to induce positive thought during mind wandering and how to inhibit negative thought during mind wandering. Doing so will contribute to improve creative problem solving without generation of negative affect.

Keywords: creativity, experience sampling method, mental health, mind wandering

Procedia PDF Downloads 155

Authors: Shivahari Revathi Venkateswaran

Abstract:

Segmenting customers plays a significant role in churn prediction. It helps the marketing team with proactive and reactive customer retention. For the reactive retention, the retention team reaches out to customers who already showed intent to disconnect by giving some special offers. When coming to proactive retention, the marketing team uses churn prediction model, which ranks each customer from rank 1 to 100, where 1 being more risk to churn/disconnect (high ranks have high propensity to churn). The churn prediction model is built by using XGBoost model. However, with the churn rank, the marketing team can only reach out to the customers based on their individual ranks. To profile different groups of customers and to frame different marketing strategies for targeted groups of customers are not possible with the churn ranks. For this, the customers must be grouped in different segments based on their profiles, like demographics and other non-controllable attributes. This helps the marketing team to frame different offer groups for the targeted audience and prevent them from disconnecting (proactive retention). For segmentation, machine learning approaches like k-mean clustering will not form unique customer segments that have customers with same attributes. This paper finds an alternate approach to find all the combination of unique segments that can be formed from the user attributes and then finds the segments who have uplift (churn rate higher than the baseline churn rate). For this, search algorithms like fast search and recursive search are used. Further, for each segment, all customers can be targeted using individual churn ranks from the churn prediction model. Finally, a UI (User Interface) is developed for the marketing team to interactively search for the meaningful segments that are formed and target the right set of audience for future marketing campaigns and prevent them from disconnecting.

Keywords: churn prediction modeling, XGBoost model, uplift segments, proactive marketing, search algorithms, retention, k-mean clustering

Procedia PDF Downloads 56

Authors: K. M. Jananni, R. Nandini

Abstract:

The proposed wireless, inverter type design of a LPG leakage monitoring system aims to provide a smart and safe kitchen. The system detects the LPG gas leak using Nano-sensors and alerts the concerned individual through GSM system. The system uses two sensors, one attached to the chimney and other to the regulator of the LPG cylinder. Upon a leakage being detected, the sensor at the regulator actuates the system to cut off the gas supply immediately using a solenoid control valve. The sensor at the chimney checks for the permissible level of LPG mix in the air and when the level exceeds the threshold, the system sends an automatic SMS to the numbers saved. Further the sensor actuates the mini suction system fixed at the chimney within 20 seconds of a leakage to suck out the gas until the level falls well below the threshold. As a safety measure, an automatic window opening and alarm feature is also incorporated into the system. The key feature of this design is that the system is provided with a special inverter designed to make the device function effectively even during power failures. In this paper, utilization of sensors in the kitchen area is discussed and this gives the proposed architecture for real time field monitoring with a PIC Micro-controller.

Keywords: nano sensors, global system for mobile communication, GSM, micro controller, inverter

Procedia PDF Downloads 456

Authors: Radhika Ranjan Roy

Abstract:

Deployment of machine learning (ML)/deep learning (DL) algorithms for cyberattack detection in operational communications networks (wireless and/or wire-line) is being delayed because of low-performance parameters (e.g., recall, precision, and f₁-score). If datasets become imbalanced, which is the usual case for communications networks, the performance tends to become worse. Complexities in handling reducing dimensions of the feature sets for increasing performance are also a huge problem. Mahalanobis algorithms have been widely applied in scientific research because Mahalanobis distance metric learning is a successful framework. In this paper, we have investigated the Mahalanobis binary classifier algorithm for increasing cyberattack detection performance over communications networks as a proof of concept. We have also found that high-dimensional information in intermediate features that are not utilized as much for classification tasks in ML/DL algorithms are the main contributor to the state-of-the-art of improved performance of the Mahalanobis method, even for imbalanced and sparse datasets. With no feature reduction, MD offers uniform results for precision, recall, and f₁-score for unbalanced and sparse NSL-KDD datasets.

Keywords: Mahalanobis distance, machine learning, deep learning, NS-KDD, local intrinsic dimensionality, chi-square, positive semi-definite, area under the curve

Procedia PDF Downloads 58

Authors: Heng Yang, Tao Luo, Yakun Zhang, Kai Wang, Wei Qin, Liang Xie, Ye Yan, Erwei Yin

Abstract:

Recent years have seen that audio-visual recognition has shown great potential in a strong noise environment. The existing method of audio-visual recognition has explored methods with ResNet and feature fusion. However, on the one hand, ResNet always occupies a large amount of memory resources, restricting the application in engineering. On the other hand, the feature merging also brings some interferences in a high noise environment. In order to solve the problems, we proposed an effective framework with bidirectional distillation. At first, in consideration of the good performance in extracting of features, we chose the light model, Efficientnet as our extractor of spatial features. Secondly, self-distillation was applied to learn more information from raw data. Finally, we proposed a bidirectional distillation in decision-level fusion. In more detail, our experimental results are based on a multi-model dataset from 24 volunteers. Eventually, the lipreading accuracy of our framework was increased by 2.3% compared with existing systems, and our framework made progress in audio-visual fusion in a high noise environment compared with the system of audio recognition without visual.

Keywords: lipreading, audio-visual, Efficientnet, distillation

Procedia PDF Downloads 111

Authors: V. Saccà, A. Sarica, F. Novellino, S. Barone, T. Tallarico, E. Filippelli, A. Granata, P. Valentino, A. Quattrone

Abstract:

The work aim was to evaluate how well Random Forest (RF) and Support Vector Machine (SVM) algorithms could support the early diagnosis of Multiple Sclerosis (MS) from resting-state functional connectivity data. In particular, we wanted to explore the ability in distinguishing between controls and patients of mean signals extracted from ICA components corresponding to 15 well-known networks. Eighteen patients with early-MS (mean-age 37.42±8.11, 9 females) were recruited according to McDonald and Polman, and matched for demographic variables with 19 healthy controls (mean-age 37.55±14.76, 10 females). MRI was acquired by a 3T scanner with 8-channel head coil: (a)whole-brain T1-weighted; (b)conventional T2-weighted; (c)resting-state functional MRI (rsFMRI), 200 volumes. Estimated total lesion load (ml) and number of lesions were calculated using LST-toolbox from the corrected T1 and FLAIR. All rsFMRIs were pre-processed using tools from the FMRIB's Software Library as follows: (1) discarding of the first 5 volumes to remove T1 equilibrium effects, (2) skull-stripping of images, (3) motion and slice-time correction, (4) denoising with high-pass temporal filter (128s), (5) spatial smoothing with a Gaussian kernel of FWHM 8mm. No statistical significant differences (t-test, p < 0.05) were found between the two groups in the mean Euclidian distance and the mean Euler angle. WM and CSF signal together with 6 motion parameters were regressed out from the time series. We applied an independent component analysis (ICA) with the GIFT-toolbox using the Infomax approach with number of components=21. Fifteen mean components were visually identified by two experts. The resulting z-score maps were thresholded and binarized to extract the mean signal of the 15 networks for each subject. Statistical and machine learning analysis were then conducted on this dataset composed of 37 rows (subjects) and 15 features (mean signal in the network) with R language. The dataset was randomly splitted into training (75%) and test sets and two different classifiers were trained: RF and RBF-SVM. We used the intrinsic feature selection of RF, based on the Gini index, and recursive feature elimination (rfe) for the SVM, to obtain a rank of the most predictive variables. Thus, we built two new classifiers only on the most important features and we evaluated the accuracies (with and without feature selection) on test-set. The classifiers, trained on all the features, showed very poor accuracies on training (RF:58.62%, SVM:65.52%) and test sets (RF:62.5%, SVM:50%). Interestingly, when feature selection by RF and rfe-SVM were performed, the most important variable was the sensori-motor network I in both cases. Indeed, with only this network, RF and SVM classifiers reached an accuracy of 87.5% on test-set. More interestingly, the only misclassified patient resulted to have the lowest value of lesion volume. We showed that, with two different classification algorithms and feature selection approaches, the best discriminant network between controls and early MS, was the sensori-motor I. Similar importance values were obtained for the sensori-motor II, cerebellum and working memory networks. These findings, in according to the early manifestation of motor/sensorial deficits in MS, could represent an encouraging step toward the translation to the clinical diagnosis and prognosis.

Keywords: feature selection, machine learning, multiple sclerosis, random forest, support vector machine

Procedia PDF Downloads 225

Authors: Hala Zaghloul, Taymoor Nazmy

Abstract:

One of the great challenges of the 21st century is to build a robot that can perceive and act within its environment and communicate with people, while also exhibiting the cognitive capabilities that lead to performance like that of people. The Pulse Coupled Neural Network, PCNN, is a relative new ANN model that derived from a neural mammal model with a great potential in the area of image processing as well as target recognition, feature extraction, speech recognition, combinatorial optimization, compressed encoding. PCNN has unique feature among other types of neural network, which make it a candid to be an important approach for perceiving in cognitive systems. This work show and emphasis on the potentials of PCNN to perform different tasks related to image processing. The main drawback or the obstacle that prevent the direct implementation of such technique, is the need to find away to control the PCNN parameters toward perform a specific task. This paper will evaluate the performance of PCNN standard model for processing images with different properties, and select the important parameters that give a significant result, also, the approaches towards find a way for the adaptation of the PCNN parameters to perform a specific task.

Keywords: cognitive system, image processing, segmentation, PCNN kernels

Procedia PDF Downloads 259

Authors: ZhenlinJu, Christopher P. Vellano, RehanAkbani, Yiling Lu, Gordon B. Mills

Abstract:

The epithelial–mesenchymal transition (EMT) is a process by which epithelial cells acquire mesenchymal characteristics, such as profound disruption of cell-cell junctions, loss of apical-basolateral polarity, and extensive reorganization of the actin cytoskeleton to induce cell motility and invasion. A hallmark of EMT is its capacity to promote metastasis, which is due in part to activation of several transcription factors and subsequent downregulation of E-cadherin. Unfortunately, current approaches have yet to uncover robust protein marker sets that can classify tumors as possessing strong EMT signatures. In this study, we utilize reverse phase protein array (RPPA) data and consensus clustering methods to successfully classify a subset of cervical squamous cell carcinoma and endocervical adenocarcinoma (CESC) tumors into an EMT protein signaling group (EMT group). The overall survival (OS) of patients in the EMT group is significantly worse than those in the other Hormone and PI3K/AKT signaling groups. In addition to a shrinkage and selection method for linear regression (LASSO), we applied training/test set and Monte Carlo resampling approaches to identify a set of protein markers that predicts the EMT status of CESC tumors. We fit a logistic model to these protein markers and developed a classifier, which was fixed in the training set and validated in the testing set. The classifier robustly predicted the EMT status of the testing set with an area under the curve (AUC) of 0.975 by Receiver Operating Characteristic (ROC) analysis. This method not only identifies a core set of proteins underlying an EMT signature in cervical cancer patients, but also provides a tool to examine protein predictors that drive molecular subtypes in other diseases.

Keywords: consensus clustering, TCGA CESC, Silhouette, Monte Carlo LASSO

Procedia PDF Downloads 445

Authors: Jiawen Wang, Qijun Chen

Abstract:

The performance of most machine-learning methods for image inpainting depends on the quantity and quality of the training samples. However, it is very expensive or even impossible to obtain a great number of training samples in many scenarios. In this paper, an image inpainting model based on a generative adversary network (GAN) is constructed for the cases when the number of training samples is small. Firstly, a feature extraction network (F-net) is incorporated into the GAN network to utilize the available information of the inpainting image. The weighted sum of the extracted feature and the random noise acts as the input to the generative network (G-net). The proposed network can be trained well even when the sample size is very small. Secondly, in the phase of the completion for each damaged image, a genetic algorithm is designed to search an optimized noise input for G-net; based on this optimized input, the parameters of the G-net and F-net are further learned (Once the completion for a certain damaged image ends, the parameters restore to its original values obtained in the training phase) to generate an image patch that not only can fill the missing part of the damaged image smoothly but also has visual semantics.

Keywords: image inpainting, generative adversary nets, genetic algorithm, small-sample size

Procedia PDF Downloads 114

Authors: Rajpal Kaur, Pooja Choudhary

Abstract:

Biometrics, which refers to identifying an individual based on his or her physiological or behavioral characteristics, has the capabilities to the reliably distinguish between an authorized person and an imposter. The Signature recognition systems can categorized as offline (static) and online (dynamic). This paper presents Surf Feature based recognition of offline signatures system that is trained with low-resolution scanned signature images. The signature of a person is an important biometric attribute of a human being which can be used to authenticate human identity. However the signatures of human can be handled as an image and recognized using computer vision and HMM techniques. With modern computers, there is need to develop fast algorithms for signature recognition. There are multiple techniques are defined to signature recognition with a lot of scope of research. In this paper, (static signature) off-line signature recognition & verification using surf feature with HMM is proposed, where the signature is captured and presented to the user in an image format. Signatures are verified depended on parameters extracted from the signature using various image processing techniques. The Off-line Signature Verification and Recognition is implemented using Mat lab platform. This work has been analyzed or tested and found suitable for its purpose or result. The proposed method performs better than the other recently proposed methods.

Keywords: offline signature verification, offline signature recognition, signatures, SURF features, HMM

Procedia PDF Downloads 364

Authors: Naser Safdarian, Nader Jafarnia Dabanloo

Abstract:

In this paper we used four features i.e. Q-wave integral, QRS complex integral, T-wave integral and total integral as extracted feature from normal and patient ECG signals to detection and localization of myocardial infarction (MI) in left ventricle of heart. In our research we focused on detection and localization of MI in standard ECG. We use the Q-wave integral and T-wave integral because this feature is important impression in detection of MI. We used some pattern recognition method such as Artificial Neural Network (ANN) to detect and localize the MI. Because these methods have good accuracy for classification of normal and abnormal signals. We used one type of Radial Basis Function (RBF) that called Probabilistic Neural Network (PNN) because of its nonlinearity property, and used other classifier such as k-Nearest Neighbors (KNN), Multilayer Perceptron (MLP) and Naive Bayes Classification. We used PhysioNet database as our training and test data. We reached over 80% for accuracy in test data for localization and over 95% for detection of MI. Main advantages of our method are simplicity and its good accuracy. Also we can improve accuracy of classification by adding more features in this method. A simple method based on using only four features which extracted from standard ECG is presented which has good accuracy in MI localization.

Keywords: ECG signal processing, myocardial infarction, features extraction, pattern recognition

Procedia PDF Downloads 436

Authors: Kay Thinzar Phu, Lwin Lwin Oo

Abstract:

In an intelligent transport system and advanced driver assistance system, the developing of real-time traffic sign detection and recognition (TSDR) system plays an important part in recent research field. There are many challenges for developing real-time TSDR system due to motion artifacts, variable lighting and weather conditions and situations of traffic signs. Researchers have already proposed various methods to minimize the challenges problem. The aim of the proposed research is to develop an efficient and effective TSDR in real time. This system proposes an adaptive thresholding method based on RGB color for traffic signs detection and new features for traffic signs recognition. In this system, the RGB color thresholding is used to detect the blue and yellow color traffic signs regions. The system performs the shape identify to decide whether the output candidate region is traffic sign or not. Lastly, new features such as termination points, bifurcation points, and 90’ angles are extracted from validated image. This system uses Myanmar Traffic Sign dataset.

Keywords: adaptive thresholding based on RGB color, blue color detection, feature extraction, yellow color detection

Procedia PDF Downloads 283

Authors: Jamshaid Ul Rahman, Faiza Makhdoom, Dianchen Lu

Abstract:

Activation functions play a decisive role in determining the capacity of Deep Neural Networks (DNNs) as they enable neural networks to capture inherent nonlinearities present in data fed to them. The prior research on activation functions primarily focused on the utility of monotonic or non-oscillatory functions, until Growing Cosine Unit (GCU) broke the taboo for a number of applications. In this paper, a Convolutional Neural Network (CNN) model named as ASU-CNN is proposed which utilizes recently designed activation function ASU across its layers. The effect of this non-monotonic and oscillatory function is inspected through feature map visualizations from different convolutional layers. The optimization of proposed network is offered by Adam with a fine-tuned adjustment of learning rate. The network achieved promising results on both training and testing data for the classification of CIFAR-10. The experimental results affirm the computational feasibility and efficacy of the proposed model for performing tasks related to the field of computer vision.

Keywords: amplifying sine unit, activation function, convolutional neural networks, oscillatory activation, image classification, CIFAR-10

Procedia PDF Downloads 83

Authors: Netanel Stern

Abstract:

Artificial intelligence (AI) algorithms have emerged as powerful tools in the field of psychiatry, offering new possibilities for enhancing diagnosis and treatment outcomes. This article explores the utilization of AI algorithms in psychiatry, highlighting their potential to revolutionize patient care. Various AI algorithms, including machine learning, natural language processing (NLP), reinforcement learning, clustering, and Bayesian networks, are discussed in detail. Moreover, ethical considerations and future directions for research and implementation are addressed.

Keywords: AI, software engineering, psychiatry, neuroimaging

Procedia PDF Downloads 83

Authors: Anderson S. Fonseca, Italo F. S. Da Silva, Robert D. A. Santos, Mayara G. Da Silva, Pedro H. C. Vieira, Antonio M. S. Sobrinho, Victor H. B. Lemos, Petterson S. Diniz, Anselmo C. Paiva, Eliana M. G. Monteiro

Abstract:

In Brazil, the National Electric Energy Agency (ANEEL) establishes that electrical energy companies are responsible for measuring and billing their customers. Among these regulations, it’s defined that a company must bill your customers within 27-33 days. If a relocation or a change of period is required, the consumer must be notified in writing, in advance of a billing period. To make it easier to organize a workday’s measurements, these companies create a reading plan. These plans consist of grouping customers into reading groups, which are visited by an employee responsible for measuring consumption and billing. The creation process of a plan efficiently and optimally is a capacitated clustering problem with constraints related to homogeneity and compactness, that is, the employee’s working load and the geographical position of the consuming unit. This process is a work done manually by several experts who have experience in the geographic formation of the region, which takes a large number of days to complete the final planning, and because it’s human activity, there is no guarantee of finding the best optimization for planning. In this paper, the GBKMeans method presents a technique based on K-Means and genetic algorithms for creating a capacitated cluster that respects the constraints established in an efficient and balanced manner, that minimizes the cost of relocating consumer units and the time required for final planning creation. The results obtained by the presented method are compared with the current planning of a real city, showing an improvement of 54.71% in the standard deviation of working load and 11.97% in the compactness of the groups.

Keywords: capacitated clustering, k-means, genetic algorithm, districting problems

Procedia PDF Downloads 179