Search results for: classifiers accuracy

Authors: Vincent Liu

Abstract:

Machine learning algorithms (ML) can achieve high accuracy in predicting outcomes compared to classical models. Metaheuristic, nature-inspired algorithms can enhance traditional ML algorithms by optimizing them such as by performing feature selection. We compare ten ML algorithms to predict 30-day hospital readmission rates for diabetes patients in the US using a dataset from UCI Machine Learning Repository with feature selection performed by Greywolf nature-inspired algorithm. The baseline accuracy for the initial random forest model was 65%. After performing feature engineering, SMOTE for class balancing, and Greywolf optimization, the machine learning algorithms showed better metrics, including F1 scores, accuracy, and confusion matrix with improvements ranging in 10%-30%, and a best model of XGBoost with an accuracy of 95%. Applying machine learning this way can improve patient outcomes as unnecessary rehospitalizations can be prevented by focusing on patients that are at a higher risk of readmission.

Keywords: diabetes, machine learning, 30-day readmission, metaheuristic

Procedia PDF Downloads 63

Authors: Jason Chen, Jennifer Chen, Shiyu Li

Abstract:

We examine whether managerial ability (MA), the passing of Sarbanes-Oxley in 2002 (SOX), and corporate operational complexity affect the accuracy of environmental capital expenditure projections of the environmentally sensitive industries (ESI). Prior studies found that firms in the ESI manipulated their projected environmental capital expenditures as a tool to achieve corporate legitimation and suggested that human factors must be examined to determine whether they are part of the determinants. We use MA to proxy for the latent human factors to examine whether MA affects the accuracy of financial disclosures in the ESI. To expand Chen and Chen (2020), we further investigate whether (1) SOX and (2) firms with complex operations and financial reporting in conjunction with MA affect firms’ projection accuracy. We find, overall, that MA is positively correlated with firm’s projection accuracy in the annual 10-Ks. Furthermore, results suggest that SOX has a positive, yet temporary, effect on MA, and that leads to better accuracy. Finally, MA matters for firms with more complex operations and financial reporting to make less projection errors than their less-complex counterparts. These results suggest that MA is a determinant that affects the accuracy of environmental capital expenditure projections for the firms in the ESI.

Keywords: managerial ability, environmentally sensitive industries, sox, corporate operational complexity

Procedia PDF Downloads 148

Authors: Gyoutae Park, Seungho Han, Byungduk Kim, Youngdo Jo, Yongsop Shim, Yeonjae Lee, Sangguk Ahn, Hiesik Kim, Jungil Park

Abstract:

In this paper, we presented not only development technology of an explosion proof type and portable combustible gas leak detector but also algorithm to improve accuracy for measuring gas concentrations. The presented techniques are to apply the flame-proof enclosure and intrinsic safe explosion proof to an infrared gas leak detector at first in Korea and to improve accuracy using linearization recursion equation and Lagrange interpolation polynomial. Together, we tested sensor characteristics and calibrated suitable input gases and output voltages. Then, we advanced the performances of combustible gaseous detectors through reflecting demands of gas safety management fields. To check performances of two company's detectors, we achieved the measurement tests with eight standard gases made by Korea Gas Safety Corporation. We demonstrated our instruments better in detecting accuracy other than detectors through experimental results.

Keywords: accuracy improvement, IR gas sensor, gas leak, detector

Procedia PDF Downloads 392

Authors: Bo Wang

Abstract:

As the first high-resolution civil optical satellite, ZY-3 satellite is able to obtain high-resolution multi-view images with three linear array sensors. The images can be used to generate Digital Elevation Models (DEM) through dense matching of stereo images. However, due to the clouds, forest, water and buildings covered on the images, there are some problems in the dense matching results such as outliers and areas failed to be matched (matching holes). This paper introduced an algorithm to verify the accuracy of DEM that generated by ZY-3 satellite with Shuttle Radar Topography Mission (SRTM). Since the accuracy of SRTM (Internal accuracy: 5 m; External accuracy: 15 m) is relatively uniform in the worldwide, it may be used to improve the accuracy of ZY-3 DEM. Based on the analysis of mass DEM and SRTM data, the processing can be divided into two aspects. The registration of ZY-3 DEM and SRTM can be firstly performed using the conjugate line features and area features matched between these two datasets. Then the ZY-3 DEM can be refined by eliminating the matching outliers and filling the matching holes. The matching outliers can be eliminated based on the statistics on Local Vector Binning (LVB). The matching holes can be filled by the elevation interpolated from SRTM. Some works are also conducted for the accuracy statistics of the ZY-3 DEM.

Keywords: ZY-3 satellite imagery, DEM, SRTM, refinement

Procedia PDF Downloads 345

Authors: Sally Salah El Menshawy, Ghada M. Ahmed GabAllah, Doaa Khedr M. Khedr

Abstract:

Background: Our study aims to detect the diagnostic role of DTI in the differentiation of salivary glands benign and malignant lesions. Results: Our study included 50 patients (25males and 25 females) divided into 4 groups (benign lesions n=20, malignant tumors n=13, post-operative changes n=10 and normal n=7). 28 patients were with parotid gland lesions, 4 patients were with submandibular gland lesions and only 1 case with sublingual gland affection. The mean fractional anisotropy (FA) and apparent diffusion coefficient (ADC) of malignant salivary gland tumors (n = 13) (0.380±0.082 and 0.877±0.234× 10⁻³ mm² s⁻¹) were significantly different (P<0.001) than that of benign tumors (n = 20) (0.147±0.03 and 1.47±0.605 × 10⁻³ mm² s⁻¹), respectively. The mean FA and ADC of post-operative changes (n = 10) were (0.211±0.069 and 1.63±0.20× 10⁻³ mm² s⁻¹) while that of normal glands (n =7) was (0.251±0.034and 1.54±0.29× 10⁻³ mm² s⁻¹), respectively. Using ADC to differentiate malignant lesions from benign lesions has an (AUC) of 0.810, with an accuracy of 69.7%. ADC used to differentiate malignant lesions from post-operative changes has (AUC) of 1.0, and an accuracy of 95.7%. FA used to discriminate malignant from benign lesions has (AUC) of 1.0, and an accuracy of 93.9%. FA used to differentiate malignant from post-operative changes has (AUC) of 0.923, and an accuracy of 95.7%. Combined FA and ADC used to differentiate malignant from benign lesions has (AUC) of 1.0, and an accuracy of 100%. Combined FA and ADC used to differentiate malignant from post-operative changes has (AUC) of 1.0, and an accuracy of 100%. Conclusion: Combined FA and ADC can differentiate malignant tumors from benign salivary gland lesions.

Keywords: diffusion tensor imaging, MRI, salivary gland, tumors

Procedia PDF Downloads 113

Authors: Jadisha Cornejo, Helio Pedrini

Abstract:

Recognition of emotions based on facial expressions has received increasing attention from the scientific community over the last years. Several fields of applications can benefit from facial emotion recognition, such as behavior prediction, interpersonal relations, human-computer interactions, recommendation systems. In this work, we develop and analyze an emotion recognition framework based on facial expressions robust to occlusions through the Weber Local Descriptor (WLD). Initially, the occluded facial expressions are reconstructed following an extension approach of Robust Principal Component Analysis (RPCA). Then, WLD features are extracted from the facial expression representation, as well as Local Binary Patterns (LBP) and Histogram of Oriented Gradients (HOG). The feature vector space is reduced using Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA). Finally, K-Nearest Neighbor (K-NN) and Support Vector Machine (SVM) classifiers are used to recognize the expressions. Experimental results on three public datasets demonstrated that the WLD representation achieved competitive accuracy rates for occluded and non-occluded facial expressions compared to other approaches available in the literature.

Keywords: emotion recognition, facial expression, occlusion, fiducial landmarks

Procedia PDF Downloads 183

Authors: Mohammadhossein Moghimi, Robert Stone, Pia Rotshtein

Abstract:

Recording psychological and physiological correlates of human performance within virtual environments and interpreting their impacts on human engagement, ‘immersion’ and related emotional or ‘effective’ states is both academically and technologically challenging. By exposing participants to an effective, real-time (game-like) virtual environment, designed and evaluated in an earlier study, a psychophysiological database containing the EEG, GSR and Heart Rate of 30 male and female gamers, exposed to 10 games, was constructed. Some 174 features were subsequently identified and extracted from a number of windows, with 28 different timing lengths (e.g. 2, 3, 5, etc. seconds). After reducing the number of features to 30, using a feature selection technique, K-Nearest Neighbour (KNN) and Support Vector Machine (SVM) methods were subsequently employed for the classification process. The classifiers categorised the psychophysiological database into four effective clusters (defined based on a 3-dimensional space – valence, arousal and dominance) and eight emotion labels (relaxed, content, happy, excited, angry, afraid, sad, and bored). The KNN and SVM classifiers achieved average cross-validation accuracies of 97.01% (±1.3%) and 92.84% (±3.67%), respectively. However, no significant differences were found in the classification process based on effective clusters or emotion labels.

Keywords: virtual reality, effective computing, effective VR, emotion-based effective physiological database

Procedia PDF Downloads 234

Authors: Belayneh Matebie, Michael Melese

Abstract:

The study focuses on addressing the challenges in classifying and ensuring the quality of Eragrostis Teff, a small and round grain that is the smallest cereal grain. Employing a traditional classification method is challenging because of its small size and the similarity of its environmental characteristics. To overcome this, this study employs a machine learning approach to develop a source and quality classification system for Teff cereal. Data is collected from various production areas in the Amhara regions, considering two types of cereal (high and low quality) across eight classes. A total of 5,920 images are collected, with 740 images for each class. Image enhancement techniques, including scaling, data augmentation, histogram equalization, and noise removal, are applied to preprocess the data. Convolutional Neural Network (CNN) is then used to extract relevant features and reduce dimensionality. The dataset is split into 80% for training and 20% for testing. Different classifiers, including FVGG16, FINCV3, QSCTC, EMQSCTC, SVM, and RF, are employed for classification, achieving accuracy rates ranging from 86.91% to 97.72%. The ensemble of FVGG16, FINCV3, and QSCTC using the Max-Voting approach outperforms individual algorithms.

Keywords: Teff, ensemble learning, max-voting, CNN, SVM, RF

Procedia PDF Downloads 57

Authors: Muhtasim Shafi Kader, Fizar Ahmed, Annesha Acharjee

Abstract:

Machine learning and data mining are crucial in health care, as well as medical information and detection. Machine learning approaches are now being utilized to improve awareness of a variety of critical health issues, including diabetes detection, neuron cell tumor diagnosis, COVID 19 identification, and so on. Parkinson’s disease is basically a disease for our senior citizens in Bangladesh. Parkinson's Disease indications often seem progressive and get worst with time. People got affected trouble walking and communicating with the condition advances. Patients can also have psychological and social vagaries, nap problems, hopelessness, reminiscence loss, and weariness. Parkinson's disease can happen in both men and women. Though men are affected by the illness at a proportion that is around partial of them are women. In this research, we have to get out the accurate ML algorithm to find out the disease with a predictable dataset and the model of the following machine learning classifiers. Therefore, nine ML classifiers are secondhand to portion study to use machine learning approaches like as follows, Naive Bayes, Adaptive Boosting, Bagging Classifier, Decision Tree Classifier, Random Forest classifier, XBG Classifier, K Nearest Neighbor Classifier, Support Vector Machine Classifier, and Gradient Boosting Classifier are used.

Keywords: naive bayes, adaptive boosting, bagging classifier, decision tree classifier, random forest classifier, XBG classifier, k nearest neighbor classifier, support vector classifier, gradient boosting classifier

Procedia PDF Downloads 130

Authors: Wei-Jong Yang, Yu-Siang Su, Pau-Choo Chung, Jar-Ferr Yang

Abstract:

Moving object detection (MOD) is an important issue in advanced driver assistance systems (ADAS). There are two important moving objects, pedestrians and scooters in ADAS. In real-world systems, there exist two important challenges for MOD, including the computational complexity and the detection accuracy. The histogram of oriented gradient (HOG) features can easily detect the edge of object without invariance to changes in illumination and shadowing. However, to reduce the execution time for real-time systems, the image size should be down sampled which would lead the outlier influence to increase. For this reason, we propose the histogram of uniformly-oriented gradient (HUG) features to get better accurate description of the contour of human body. In the testing phase, the support vector machine (SVM) with linear kernel function is involved. Experimental results show the correctness and effectiveness of the proposed method. With SVM classifiers, the real testing results show the proposed HUG features achieve better than classification performance than the HOG ones.

Keywords: moving object detection, histogram of oriented gradient, histogram of uniformly-oriented gradient, linear support vector machine

Procedia PDF Downloads 595

Authors: Mário Ernesto Sitoe, Orlando Zacarias

Abstract:

University education is influenced by several factors that range from the adoption of strategies to strengthen the whole process to the academic performance improvement of the students themselves. This work uses data mining techniques to develop a predictive model to identify students with a tendency to evasion and retention. To this end, a database of real students’ data from the Department of University Admission (DAU) and the Department of Mathematics and Informatics (DMI) was used. The data comprised 388 undergraduate students admitted in the years 2009 to 2014. The Weka tool was used for model building, using three different techniques, namely: K-nearest neighbor, random forest, and logistic regression. To allow for training on multiple train-test splits, a cross-validation approach was employed with a varying number of folds. To reduce bias variance and improve the performance of the models, ensemble methods of Bagging and Stacking were used. After comparing the results obtained by the three classifiers, Logistic Regression using Bagging with seven folds obtained the best performance, showing results above 90% in all evaluated metrics: accuracy, rate of true positives, and precision. Retention is the most common tendency.

Keywords: evasion and retention, cross-validation, bagging, stacking

Procedia PDF Downloads 84

Authors: Mohammad Karimi Moridani, Mohammad Abdi Zadeh, Zahra Shahiazar Mazraeh

Abstract:

In recent years, the use of intelligent systems in biomedical engineering has increased dramatically, especially in the diagnosis of various diseases. Also, due to the relatively simple recording of the electrocardiogram signal (ECG), this signal is a good tool to show the function of the heart and diseases associated with it. The aim of this paper is to design an intelligent system for automatically detecting a normal electrocardiogram signal from abnormal one. Using this diagnostic system, it is possible to identify a person's heart condition in a very short time and with high accuracy. The data used in this article are from the Physionet database, available in 2016 for use by researchers to provide the best method for detecting normal signals from abnormalities. Data is of both genders and the data recording time varies between several seconds to several minutes. All data is also labeled normal or abnormal. Due to the low positional accuracy and ECG signal time limit and the similarity of the signal in some diseases with the normal signal, the heart rate variability (HRV) signal was used. Measuring and analyzing the heart rate variability with time to evaluate the activity of the heart and differentiating different types of heart failure from one another is of interest to the experts. In the preprocessing stage, after noise cancelation by the adaptive Kalman filter and extracting the R wave by the Pan and Tampkinz algorithm, R-R intervals were extracted and the HRV signal was generated. In the process of processing this paper, a new idea was presented that, in addition to using the statistical characteristics of the signal to create a return map and extraction of nonlinear characteristics of the HRV signal due to the nonlinear nature of the signal. Finally, the artificial neural networks widely used in the field of ECG signal processing as well as distinctive features were used to classify the normal signals from abnormal ones. To evaluate the efficiency of proposed classifiers in this paper, the area under curve ROC was used. The results of the simulation in the MATLAB environment showed that the AUC of the MLP and SVM neural network was 0.893 and 0.947, respectively. As well as, the results of the proposed algorithm in this paper indicated that the more use of nonlinear characteristics in normal signal classification of the patient showed better performance. Today, research is aimed at quantitatively analyzing the linear and non-linear or descriptive and random nature of the heart rate variability signal, because it has been shown that the amount of these properties can be used to indicate the health status of the individual's heart. The study of nonlinear behavior and dynamics of the heart's neural control system in the short and long-term provides new information on how the cardiovascular system functions, and has led to the development of research in this field. Given that the ECG signal contains important information and is one of the common tools used by physicians to diagnose heart disease, but due to the limited accuracy of time and the fact that some information about this signal is hidden from the viewpoint of physicians, the design of the intelligent system proposed in this paper can help physicians with greater speed and accuracy in the diagnosis of normal and patient individuals and can be used as a complementary system in the treatment centers.

Keywords: neart rate variability, signal processing, linear and non-linear features, classification methods, ROC Curve

Procedia PDF Downloads 264

Authors: Zachary Pan

Abstract:

Our general mission is to solve the problem of classifying images into different body part types and deciding if each of them is healthy or not. However, for now, we will determine healthiness for only one-sixth of the body parts, specifically the chest. We will detect pneumonia in X-ray scans of those chest images. With this type of AI, doctors can use it as a second opinion when they are taking CT or X-ray scans of their patients. Another ad-vantage of using this machine learning classifier is that it has no human weaknesses like fatigue. The overall ap-proach to this problem is to split the problem into two parts: first, classify the image, then determine if it is healthy. In order to classify the image into a specific body part class, the body parts dataset must be split into test and training sets. We can then use many models, like neural networks or logistic regression models, and fit them using the training set. Now, using the test set, we can obtain a realistic accuracy the models will have on images in the real world since these testing images have never been seen by the models before. In order to increase this testing accuracy, we can also apply many complex algorithms to the models, like multiplicative weight update. For the second part of the problem, to determine if the body part is healthy, we can have another dataset consisting of healthy and non-healthy images of the specific body part and once again split that into the test and training sets. We then use another neural network to train on those training set images and use the testing set to figure out its accuracy. We will do this process only for the chest images. A major conclusion reached is that convolutional neural networks are the most reliable and accurate at image classification. In classifying the images, the logistic regression model, the neural network, neural networks with multiplicative weight update, neural networks with the black box algorithm, and the convolutional neural network achieved 96.83 percent accuracy, 97.33 percent accuracy, 97.83 percent accuracy, 96.67 percent accuracy, and 98.83 percent accuracy, respectively. On the other hand, the overall accuracy of the model that de-termines if the images are healthy or not is around 78.37 percent accuracy.

Keywords: body part, healthcare, machine learning, neural networks

Procedia PDF Downloads 109

Authors: Swati Kishor Zode, Rahul Ambekar

Abstract:

Data mining is the extraction of fascinating examples on the other hand information from enormous measure of information and choice is made as indicated by the applicable information extracted. As of late, with the dangerous advancement in internet, stockpiling of information and handling procedures, privacy preservation has been one of the major (higher) concerns in data mining. Various techniques and methods have been produced for protection saving data mining. In the situation of Clinical Decision Support System, the choice is to be made on the premise of the data separated from the remote servers by means of Internet to diagnose the patient. In this paper, the fundamental thought is to build the precision of Decision Support System for multiple diseases for different maladies and in addition protect persistent information while correspondence between Clinician side (Client side) also, the Server side. A privacy preserving protocol for clinical decision support network is proposed so that patients information dependably stay scrambled amid diagnose prepare by looking after the accuracy. To enhance the precision of Decision Support System for various malady C5.0 classifiers and to save security, a Homomorphism encryption algorithm Paillier cryptosystem is being utilized.

Keywords: classification, homomorphic encryption, clinical decision support, privacy

Procedia PDF Downloads 330

Authors: Jung Ae Park, Jooyong Park

Abstract:

Writing is an important learning activity that cultivates higher level thinking. Effective and immediate feedback is necessary to help improve students' writing skills. Peer assessment can be an effective method in writing tasks because it makes it possible for students not only to receive quick feedback on their writing but also to get a chance to examine different perspectives on the same topic. Peer assessment can be practiced frequently and has the advantage of immediate feedback. However, there is controversy about the accuracy of peer assessment. In this study, we tried to demonstrate experimentally how the accuracy of peer assessment could be improved. Participants (n=76) were randomly assigned to groups of 4 members. All the participant graded two sets of 4 essays on the same topic. They graded the first set twice, and the second set or the posttest once. After the first grading of the first set, each group in the experimental condition 1 (discussion group), were asked to discuss the results of the peer assessment and then to grade the essays again. Each group in the experimental condition 2 (reading group), were asked to read the assessment on each essay by an expert and then to grade the essays again. In the control group, the participants were asked to grade the 4 essays twice in different orders. Afterwards, all the participants graded the second set of 4 essays. The mean score from 4 participants was calculated for each essay. The accuracy of the peer assessment was measured by Pearson correlation with the scores of the expert. The results were analyzed by two-way repeated measure ANOVA. The main effect of grading was observed: Grading accuracy got better as the number of grading experience increased. Analysis of posttest accuracy revealed that the score variations within a group of 4 participants decreased in both discussion and reading conditions but not in the control condition. These results suggest that having students discuss their grading together can be an efficient means to improve peer assessment accuracy. By discussing, students can learn from others about what to consider in grading and whether their grading is too strict or lenient. Further research is needed to examine the exact cause of the grading accuracy.

Keywords: peer assessment, evaluation accuracy, discussion, score variations

Procedia PDF Downloads 267

Authors: Lakshmidevi Sreeramareddy, Komalpreet Kaur, Nane Pothier

Abstract:

Graphical-based passwords have existed for decades. Their major advantage is that they are easier to remember than an alphanumeric password. However, their disadvantage (especially recognition-based passwords) is the smaller password space, making them more vulnerable to brute force attacks. Graphical passwords are also highly susceptible to the shoulder-surfing effect. The gesture-based password method that we developed is a grid-free, template-free method. In this study, we evaluated the gesture-based passwords for usability and vulnerability. The results of the study are significant. We developed a gesture-based password application for data collection. Two modes of data collection were used: Creation mode and Replication mode. In creation mode (Session 1), users were asked to create six different passwords and reenter each password five times. In replication mode, users saw a password image created by some other user for a fixed duration of time. Three different duration timers, such as 5 seconds (Session 2), 10 seconds (Session 3), and 15 seconds (Session 4), were used to mimic the shoulder-surfing attack. After the timer expired, the password image was removed, and users were asked to replicate the password. There were 74, 57, 50, and 44 users participated in Session 1, Session 2, Session 3, and Session 4 respectfully. In this study, the machine learning algorithms have been applied to determine whether the person is a genuine user or an imposter based on the password entered. Five different machine learning algorithms were deployed to compare the performance in user authentication: namely, Decision Trees, Linear Discriminant Analysis, Naive Bayes Classifier, Support Vector Machines (SVMs) with Gaussian Radial Basis Kernel function, and K-Nearest Neighbor. Gesture-based password features vary from one entry to the next. It is difficult to distinguish between a creator and an intruder for authentication. For each password entered by the user, four features were extracted: password score, password length, password speed, and password size. All four features were normalized before being fed to a classifier. Three different classifiers were trained using data from all four sessions. Classifiers A, B, and C were trained and tested using data from the password creation session and the password replication with a timer of 5 seconds, 10 seconds, and 15 seconds, respectively. The classification accuracies for Classifier A using five ML algorithms are 72.5%, 71.3%, 71.9%, 74.4%, and 72.9%, respectively. The classification accuracies for Classifier B using five ML algorithms are 69.7%, 67.9%, 70.2%, 73.8%, and 71.2%, respectively. The classification accuracies for Classifier C using five ML algorithms are 68.1%, 64.9%, 68.4%, 71.5%, and 69.8%, respectively. SVMs with Gaussian Radial Basis Kernel outperform other ML algorithms for gesture-based password authentication. Results confirm that the shorter the duration of the shoulder-surfing attack, the higher the authentication accuracy. In conclusion, behavioral features extracted from the gesture-based passwords lead to less vulnerable user authentication.

Keywords: authentication, gesture-based passwords, machine learning algorithms, shoulder-surfing attacks, usability

Procedia PDF Downloads 107

Authors: Selvakumar Ulaganathan, Ivo Couckuyt, Francesco Ferranti, Tom Dhaene, Eric Laermans

Abstract:

Variable-fidelity surrogate modelling offers an efficient way to approximate function data available in multiple degrees of accuracy each with varying computational cost. In this paper, a Kriging-based variable-fidelity surrogate modelling approach is introduced to approximate such deterministic data. Initially, individual Kriging surrogate models, which are enhanced with gradient data of different degrees of accuracy, are constructed. Then these Gradient enhanced Kriging surrogate models are strategically coupled using a recursive CoKriging formulation to provide an accurate surrogate model for the highest fidelity data. While, intuitively, gradient data is useful to enhance the accuracy of surrogate models, the primary motivation behind this work is to investigate if it is also worthwhile incorporating gradient data of varying degrees of accuracy.

Keywords: Kriging, CoKriging, Surrogate modelling, Variable- fidelity modelling, Gradients

Procedia PDF Downloads 558

Authors: Monica Gonzalez Machorro

Abstract:

Dementia is hard to diagnose because of the lack of early physical symptoms. Early dementia recognition is key to improving the living condition of patients. Speech technology is considered a valuable biomarker for this challenge. Recent works have utilized conventional acoustic features and machine learning methods to detect dementia in speech. BERT-like classifiers have reported the most promising performance. One constraint, nonetheless, is that these studies are either based on human transcripts or on transcripts produced by automatic speech recognition (ASR) systems. This research contribution is to explore a method that does not require transcriptions to detect early Alzheimer’s disease (AD) and mild cognitive impairment (MCI). This is achieved by fine-tuning a pre-trained ASR model for the downstream early AD and MCI tasks. To do so, a subset of the thoroughly studied Pitt Corpus is customized. The subset is balanced for class, age, and gender. Data processing also involves cropping the samples into 10-second segments. For comparison purposes, a baseline model is defined by training and testing a Random Forest with 20 extracted acoustic features using the librosa library implemented in Python. These are: zero-crossing rate, MFCCs, spectral bandwidth, spectral centroid, root mean square, and short-time Fourier transform. The baseline model achieved a 58% accuracy. To fine-tune HuBERT as a classifier, an average pooling strategy is employed to merge the 3D representations from audio into 2D representations, and a linear layer is added. The pre-trained model used is ‘hubert-large-ls960-ft’. Empirically, the number of epochs selected is 5, and the batch size defined is 1. Experiments show that our proposed method reaches a 69% balanced accuracy. This suggests that the linguistic and speech information encoded in the self-supervised ASR-based model is able to learn acoustic cues of AD and MCI.

Keywords: automatic speech recognition, early Alzheimer’s recognition, mild cognitive impairment, speech impairment

Procedia PDF Downloads 127

Authors: Elia Favarelli, Enrico Testi, Andrea Giorgetti

Abstract:

Autonomous structural health monitoring (SHM) of many structures and bridges became a topic of paramount importance for maintenance purposes and safety reasons. This paper proposes a set of machine learning (ML) tools to perform automatic feature selection and detection of anomalies in a bridge from vibrational data and compare different feature extraction schemes to increase the accuracy and reduce the amount of data collected. As a case study, the Z-24 bridge is considered because of the extensive database of accelerometric data in both standard and damaged conditions. The proposed framework starts from the first four fundamental frequencies extracted through operational modal analysis (OMA) and clustering, followed by density-based time-domain filtering (tracking). The fundamental frequencies extracted are then fed to a dimensionality reduction block implemented through two different approaches: feature selection (intelligent multiplexer) that tries to estimate the most reliable frequencies based on the evaluation of some statistical features (i.e., mean value, variance, kurtosis), and feature extraction (auto-associative neural network (ANN)) that combine the fundamental frequencies to extract new damage sensitive features in a low dimensional feature space. Finally, one class classifier (OCC) algorithms perform anomaly detection, trained with standard condition points, and tested with normal and anomaly ones. In particular, a new anomaly detector strategy is proposed, namely one class classifier neural network two (OCCNN2), which exploit the classification capability of standard classifiers in an anomaly detection problem, finding the standard class (the boundary of the features space in normal operating conditions) through a two-step approach: coarse and fine boundary estimation. The coarse estimation uses classics OCC techniques, while the fine estimation is performed through a feedforward neural network (NN) trained that exploits the boundaries estimated in the coarse step. The detection algorithms vare then compared with known methods based on principal component analysis (PCA), kernel principal component analysis (KPCA), and auto-associative neural network (ANN). In many cases, the proposed solution increases the performance with respect to the standard OCC algorithms in terms of F1 score and accuracy. In particular, by evaluating the correct features, the anomaly can be detected with accuracy and an F1 score greater than 96% with the proposed method.

Keywords: anomaly detection, frequencies selection, modal analysis, neural network, sensor network, structural health monitoring, vibration measurement

Procedia PDF Downloads 124

Authors: Benbiao Song, Yan Gao, Zhuo Liu

Abstract:

Geostatistical modeling is the key technic for reservoir characterization, the quality of geological models will influence the prediction of reservoir performance greatly, but few studies have been done to quantify the factors impacting geostatistical reservoir modeling accuracy. In this study, 16 fluvial prototype models have been established to represent different geological complexity, 6 cases range from 16 to 361 wells were defined to reproduce all those 16 prototype models by different methodologies including SIS, object-based and MPFS algorithms accompany with different constraint parameters. Modeling accuracy ratio was defined to quantify the influence of each factor, and ten realizations were averaged to represent each accuracy ratio under the same modeling condition and parameters association. Totally 5760 simulations were done to quantify the relative contribution of each factor to the simulation accuracy, and the results can be used as strategy guide for facies modeling in the similar condition. It is founded that data density, geological trend and geological complexity have great impact on modeling accuracy. Modeling accuracy may up to 90% when channel sand width reaches up to 1.5 times of well space under whatever condition by SIS and MPFS methods. When well density is low, the contribution of geological trend may increase the modeling accuracy from 40% to 70%, while the use of proper variogram may have very limited contribution for SIS method. It can be implied that when well data are dense enough to cover simple geobodies, few efforts were needed to construct an acceptable model, when geobodies are complex with insufficient data group, it is better to construct a set of robust geological trend than rely on a reliable variogram function. For object-based method, the modeling accuracy does not increase obviously as SIS method by the increase of data density, but kept rational appearance when data density is low. MPFS methods have the similar trend with SIS method, but the use of proper geological trend accompany with rational variogram may have better modeling accuracy than MPFS method. It implies that the geological modeling strategy for a real reservoir case needs to be optimized by evaluation of dataset, geological complexity, geological constraint information and the modeling objective.

Keywords: fluvial facies, geostatistics, geological trend, modeling strategy, modeling accuracy, variogram

Procedia PDF Downloads 264

Authors: Htay Htay Yi, Zin May Aye

Abstract:

Network security is role of the ICT environment because malicious users are continually growing that realm of education, business, and then related with ICT. The network security contravention is typically described and examined centrally based on a security event management system. The firewalls, Intrusion Detection System (IDS), and Intrusion Prevention System are becoming essential to monitor or prevent of potential violations, incidents attack, and imminent threats. In this system, the firewall rules are set only for where the system policies are needed. Dataset deployed in this system are derived from the testbed environment. The traffic as in DoS and PortScan traffics are applied in the testbed with firewall and IDS implementation. The network traffics are classified as normal or attacks in the existing testbed environment based on six machine learning classification methods applied in the system. It is required to be tested to get datasets and applied for DoS and PortScan. The dataset is based on CICIDS2017 and some features have been added. This system tested 26 features from the applied dataset. The system is to reduce false positive rates and to improve accuracy in the implemented testbed design. The system also proves good performance by selecting important features and comparing existing a dataset by machine learning classifiers.

Keywords: false negative rate, intrusion detection system, machine learning methods, performance

Procedia PDF Downloads 118

Authors: A. Karimzad Ghavidel, M. Zadshakouyan

Abstract:

Laser cutting is a very common production method for cutting 2D polymeric parts. Developing of polymer composites with nano-fibers makes important their other properties like laser workability. The aim of this research is investigation of the influence different laser cutting conditions on the dimensional accuracy of parts and holes from poly methyl methacrylate (PMMA)/carbon nanotubes (CNTs) material. Experiments were carried out by considering of CNTs (in four level 0,0.5, 1 and 1.5% wt.%), laser power (60, 80, and 100 watt) and cutting speed 20, 30, and 40 mm/s as input variable factors. The results reveal that CNTs adding improves the laser workability of PMMA and the increasing of power has a significant effect on the part and hole size. The findings also show cutting speed is effective parameter on the size accuracy. Eventually, the statistical analysis of results was done, and calculated mathematical equations by the regression are presented for determining relation between input and output factor.

Keywords: dimensional accuracy, PMMA, CNTs, laser cutting

Procedia PDF Downloads 307

Authors: Tokihiko Akita, Seiichi Mita

Abstract:

A millimeter-wave radar is the most robust against adverse environments, making it an essential environment recognition sensor for automated driving. However, the reflection signal is sparse and unstable, so it is difficult to obtain the high recognition accuracy. Deep learning provides high accuracy even for them in recognition, but requires large scale datasets with ground truth. Specially, it takes a lot of cost to annotate for a millimeter-wave radar. For the solution, utilizing a simulator that can generate an annotated huge dataset is effective. Simulation of the radar is more difficult to match with real world data than camera image, and recognition by deep learning with higher-order features using the simulator causes further deviation. We have challenged to improve the accuracy of traffic participant classification by fusing simulator and real-world data with domain adaptation technique. Experimental results with the domain adaptation network created by us show that classification accuracy can be improved even with a few real-world data.

Keywords: millimeter-wave radar, object classification, deep learning, simulation, domain adaptation

Procedia PDF Downloads 94

Authors: Baha Eddine Aissou, Aichouche Belhadj Aissa

Abstract:

Light detection and ranging (LiDAR) is an active remote sensing technology used for several applications. Airborne LiDAR is becoming an important technology for the acquisition of a highly accurate dense point cloud. A classification of airborne laser scanning (ALS) point cloud is a very important task that still remains a real challenge for many scientists. Support vector machine (SVM) is one of the most used statistical learning algorithms based on kernels. SVM is a non-parametric method, and it is recommended to be used in cases where the data distribution cannot be well modeled by a standard parametric probability density function. Using a kernel, it performs a robust non-linear classification of samples. Often, the data are rarely linearly separable. SVMs are able to map the data into a higher-dimensional space to become linearly separable, which allows performing all the computations in the original space. This is one of the main reasons that SVMs are well suited for high-dimensional classification problems. Only a few training samples, called support vectors, are required. SVM has also shown its potential to cope with uncertainty in data caused by noise and fluctuation, and it is computationally efficient as compared to several other methods. Such properties are particularly suited for remote sensing classification problems and explain their recent adoption. In this poster, the SVM classification of ALS LiDAR data is proposed. Firstly, connected component analysis is applied for clustering the point cloud. Secondly, the resulting clusters are incorporated in the SVM classifier. Radial basic function (RFB) kernel is used due to the few numbers of parameters (C and γ) that needs to be chosen, which decreases the computation time. In order to optimize the classification rates, the parameters selection is explored. It consists to find the parameters (C and γ) leading to the best overall accuracy using grid search and 5-fold cross-validation. The exploited LiDAR point cloud is provided by the German Society for Photogrammetry, Remote Sensing, and Geoinformation. The ALS data used is characterized by a low density (4-6 points/m²) and is covering an urban area located in residential parts of the city Vaihingen in southern Germany. The class ground and three other classes belonging to roof superstructures are considered, i.e., a total of 4 classes. The training and test sets are selected randomly several times. The obtained results demonstrated that a parameters selection can orient the selection in a restricted interval of (C and γ) that can be further explored but does not systematically lead to the optimal rates. The SVM classifier with hyper-parameters is compared with the most used classifiers in literature for LiDAR data, random forest, AdaBoost, and decision tree. The comparison showed the superiority of the SVM classifier using parameters selection for LiDAR data compared to other classifiers.

Keywords: classification, airborne LiDAR, parameters selection, support vector machine

Procedia PDF Downloads 148

Authors: Ganesh B. Shinde, Vijaya B. Musande

Abstract:

Cotton crop identification based on the timely information has significant advantage to the different implications of food, economic and environment. Due to the significant advantages, the accurate detection of cotton crop regions using supervised learning procedure is challenging problem in remote sensing. Here, classifiers on the direct image are played a major role but the results are not much satisfactorily. In order to further improve the effectiveness, variety of vegetation indices are proposed in the literature. But, recently, the major challenge is to find the better vegetation indices for the cotton crop identification through the proposed methodology. Accordingly, fuzzy c-means clustering is combined with neural network algorithm, trained by Levenberg-Marquardt for cotton crop classification. To experiment the proposed method, five LISS-III satellite images was taken and the experimentation was done with six vegetation indices such as Simple Ratio, Normalized Difference Vegetation Index, Enhanced Vegetation Index, Green Atmospherically Resistant Vegetation Index, Wide-Dynamic Range Vegetation Index, Green Chlorophyll Index. Along with these indices, Green Leaf Area Index is also considered for investigation. From the research outcome, Green Atmospherically Resistant Vegetation Index outperformed with all other indices by reaching the average accuracy value of 95.21%.

Keywords: Fuzzy C-Means clustering (FCM), neural network, Levenberg-Marquardt (LM) algorithm, vegetation indices

Procedia PDF Downloads 319

Authors: Pegah Safavi, Mehdi Zehtabian, Mohammad Amin Mosleh-Shirazi

Abstract:

Cancer is one of the leading causes of death in the world. Radiation therapy is one of the main choices for cancer treatment. Intensity-modulated radiation therapy is a new type of radiation therapy technique available for vital structures such as the parathyroid glands. It is very important to check the accuracy of the delivered IMRT treatment because any mistake may lead to more complications for the patient. This paper describes an experiment to determine the accuracy of a dose measured by EBT3 film. To test this method, the EBT3 film on the head and neck of the Rando phantom was irradiated by an IMRT device and the irradiation was repeated twice. Finally, the dose designed by the irradiation system was compared with the dose measured by the EBT3 film. Using this criterion, the accuracy of the EBT3 film was evaluated. When using this criterion, a 95% agreement was reached between the planned treatment and the measured values.

Keywords: EBT3, phantom, accuracy, cancer, IMRT

Procedia PDF Downloads 150

Authors: Aqsa Ali, Aleem Mushtaq, Attaullah Memon, Monna

Abstract:

In this paper, we present a low cost design for a smart glove that can perform sign language recognition to assist the speech impaired people. Specifically, we have designed and developed an Assistive Hand Gesture Interpreter that recognizes hand movements relevant to the American Sign Language (ASL) and translates them into text for display on a Thin-Film-Transistor Liquid Crystal Display (TFT LCD) screen as well as synthetic speech. Linear Bayes Classifiers and Multilayer Neural Networks have been used to classify 11 feature vectors obtained from the sensors on the glove into one of the 27 ASL alphabets and a predefined gesture for space. Three types of features are used; bending using six bend sensors, orientation in three dimensions using accelerometers and contacts at vital points using contact sensors. To gauge the performance of the presented design, the training database was prepared using five volunteers. The accuracy of the current version on the prepared dataset was found to be up to 99.3% for target user. The solution combines electronics, e-textile technology, sensor technology, embedded system and machine learning techniques to build a low cost wearable glove that is scrupulous, elegant and portable.

Keywords: American sign language, assistive hand gesture interpreter, human-machine interface, machine learning, sensing glove

Procedia PDF Downloads 303

Authors: Abdoulaye Diop, Nahid Emad, Thierry Winter, Mohamed Hilia

Abstract:

Data assets protection is a crucial issue in the cybersecurity field. Companies use logical access control tools to vault their information assets and protect them against external threats, but they lack solutions to counter insider threats. Nowadays, insider threats are the most significant concern of security analysts. They are mainly individuals with legitimate access to companies information systems, which use their rights with malicious intents. In several fields, behavior anomaly detection is the method used by cyber specialists to counter the threats of user malicious activities effectively. In this paper, we present the step toward the construction of a user and entity behavior analysis framework by proposing a behavior anomaly detection model. This model combines machine learning classification techniques and graph-based methods, relying on linear algebra and parallel computing techniques. We show the utility of an ensemble learning approach in this context. We present some detection methods tests results on an representative access control dataset. The use of some explored classifiers gives results up to 99% of accuracy.

Keywords: cybersecurity, data protection, access control, insider threat, user behavior analysis, ensemble learning, high performance computing

Procedia PDF Downloads 128

Authors: Deepika Gupta, Namita Bhutani, Mithlesh Bhargav

Abstract:

Objective: Breast cancer is the most commonly encountered lesion in females after non-melanoma skin malignancies. The triple assessment is an important approach in pre-operative lesions in developing countries. The objectives of the present study were to determine diagnostic accuracy and calculation of ROM along with cyto-histopathological correlation with the incorporation of a new IAC reporting system. Material and Methods: A total of 940 FNAC slides were retrieved from December 2019 to December 2020 and categorized according to the new IAC system. The diagnostic accuracy and calculation of ROM, along with cyto-histopathological correlation, were determined. Results: All the breast FNAC lesions were categorized from C1 to C5. Of the 940 cases, 358 cases had cyto-histopathological correlation. The ROM was ranging from 0% to 99%. All the statistical parameters were calculated along with diagnostic accuracy which was 97%. Conclusion: The new IAC standardized reporting system of breast FNAC evoked the utilization of rapid, accurate, and low-cost diagnostic tests and broadened the understanding and application of breast FNAC.

Keywords: accuracy, fine needle aspiration cytology, IAC, risk of malignancy, Yokohama system

Procedia PDF Downloads 11

Authors: Sarah Beaver, Renee Bryce

Abstract:

Audiograms detect hearing impairment, but missing values pose problems. This work explores imputations in an attempt to improve accuracy. This work implements Linear Regression, Lasso, Linear Support Vector Regression, Bayesian Ridge, K Nearest Neighbors (KNN), and Random Forest machine learning techniques to impute audiogram frequencies ranging from 125Hz to 8000Hz. The data contains patients who had or were candidates for cochlear implants. Accuracy is compared across two different Nested Cross-Validation k values. Over 4000 audiograms were used from 800 unique patients. Additionally, training on data combines and compares left and right ear audiograms versus single ear side audiograms. The accuracy achieved using Root Mean Square Error (RMSE) values for the best models for Random Forest ranges from 4.74 to 6.37. The R\textsuperscript{2} values for the best models for Random Forest ranges from .91 to .96. The accuracy achieved using RMSE values for the best models for KNN ranges from 5.00 to 7.72. The R\textsuperscript{2} values for the best models for KNN ranges from .89 to .95. The best imputation models received R\textsuperscript{2} between .89 to .96 and RMSE values less than 8dB. We also show that the accuracy of classification predictive models performed better with our best imputation models versus constant imputations by a two percent increase.

Keywords: machine learning, audiograms, data imputations, single imputations

Procedia PDF Downloads 83