Search results for: ensemble classifier by voting

Authors: Shahin Mirshekari, Mohammadreza Moradi, Hossein Jafari, Mehdi Jafari, Mohammad Ensaf

Abstract:

This research employs Gaussian Process Regression (GPR) with an ensemble kernel, integrating Exponential Squared, Revised Matern, and Rational Quadratic kernels to analyze pharmaceutical sales data. Bayesian optimization was used to identify optimal kernel weights: 0.76 for Exponential Squared, 0.21 for Revised Matern, and 0.13 for Rational Quadratic. The ensemble kernel demonstrated superior performance in predictive accuracy, achieving an R² score near 1.0, and significantly lower values in MSE, MAE, and RMSE. These findings highlight the efficacy of ensemble kernels in GPR for predictive analytics in complex pharmaceutical sales datasets.

Keywords: Gaussian process regression, ensemble kernels, bayesian optimization, pharmaceutical sales analysis, time series forecasting, data analysis

Procedia PDF Downloads 29

Authors: A. Sopharak, B. Uyyanonvara, S. Barman

Abstract:

Diabetic retinopathy is characterized by the development of retinal microaneurysms. The damage can be prevented if disease is treated in its early stages. In this paper, we are comparing Support Vector Machine (SVM) and Naïve Bayes (NB) classifiers for automatic microaneurysm detection in images acquired through non-dilated pupils. The Nearest Neighbor classifier is used as a baseline for comparison. Detected microaneurysms are validated with expert ophthalmologists’ hand-drawn ground-truths. The sensitivity, specificity, precision and accuracy of each method are also compared.

Keywords: diabetic retinopathy, microaneurysm, naive Bayes classifier, SVM classifier

Procedia PDF Downloads 304

Authors: Rodolfo Lorbieski, Silvia Modesto Nassar

Abstract:

Many supervised machine learning tasks require decision making across numerous different classes. Multi-class classification has several applications, such as face recognition, text recognition and medical diagnostics. The objective of this article is to analyze an adapted method of Stacking in multi-class problems, which combines ensembles within the ensemble itself. For this purpose, a training similar to Stacking was used, but with three levels, where the final decision-maker (level 2) performs its training by combining outputs from the tree-based pair of meta-classifiers (level 1) from Bayesian families. These are in turn trained by pairs of base classifiers (level 0) of the same family. This strategy seeks to promote diversity among the ensembles forming the meta-classifier level 2. Three performance measures were used: (1) accuracy, (2) area under the ROC curve, and (3) time for three factors: (a) datasets, (b) experiments and (c) levels. To compare the factors, ANOVA three-way test was executed for each performance measure, considering 5 datasets by 25 experiments by 3 levels. A triple interaction between factors was observed only in time. The accuracy and area under the ROC curve presented similar results, showing a double interaction between level and experiment, as well as for the dataset factor. It was concluded that level 2 had an average performance above the other levels and that the proposed method is especially efficient for multi-class problems when compared to binary problems.

Keywords: stacking, multi-layers, ensemble, multi-class

Procedia PDF Downloads 235

Authors: Jaehong Yu, Seoung Bum Kim

Abstract:

Clustering analysis can facilitate the extraction of intrinsic patterns in a dataset and reveal its natural groupings without requiring class information. For effective clustering analysis in high dimensional datasets, unsupervised dimensionality reduction is an important task. Unsupervised dimensionality reduction can generally be achieved by feature extraction or feature selection. In many situations, feature selection methods are more appropriate than feature extraction methods because of their clear interpretation with respect to the original features. The unsupervised feature selection can be categorized as feature subset selection and feature ranking method, and we focused on unsupervised feature ranking methods which evaluate the features based on their importance scores. Recently, several unsupervised feature ranking methods were developed based on ensemble approaches to achieve their higher accuracy and stability. However, most of the ensemble-based feature ranking methods require the true number of clusters. Furthermore, these algorithms evaluate the feature importance depending on the ensemble clustering solution, and they produce undesirable evaluation results if the clustering solutions are inaccurate. To address these limitations, we proposed an ensemble-based feature ranking method with random subspace and multiple-k ensemble (FRRM). The proposed FRRM algorithm evaluates the importance of each feature with the random subspace ensemble, and all evaluation results are combined with the ensemble importance scores. Moreover, FRRM does not require the determination of the true number of clusters in advance through the use of the multiple-k ensemble idea. Experiments on various benchmark datasets were conducted to examine the properties of the proposed FRRM algorithm and to compare its performance with that of existing feature ranking methods. The experimental results demonstrated that the proposed FRRM outperformed the competitors.

Keywords: clustering analysis, multiple-k ensemble, random subspace-based feature evaluation, unsupervised feature ranking

Procedia PDF Downloads 308

Authors: Eui-Jin Kim, Hasik Lee, Su-Jin Park, Dong-Kyu Kim

Abstract:

Smart-card data are expected to provide information on activity pattern as an alternative to conventional person trip surveys. The focus of this study is to propose a method for training the person trip surveys to supplement the smart-card data that does not contain the purpose of each trip. We selected only available features from smart card data such as spatiotemporal information on the trip and geographic information system (GIS) data near the stations to train the survey data. XGboost, which is state-of-the-art tree-based ensemble classifier, was used to train data from multiple sources. This classifier uses a more regularized model formalization to control the over-fitting and show very fast execution time with well-performance. The validation results showed that proposed method efficiently estimated the trip purpose. GIS data of station and duration of stay at the destination were significant features in modeling trip purpose.

Keywords: activity pattern, data fusion, smart-card, XGboost

Procedia PDF Downloads 214

Authors: Magdalena Musiał-Karg

Abstract:

The expansion of telecommunication and progress of electronic media constitute important elements of our times. The recent worldwide convergence of information and communication technologies (ICT) and dynamic development of the mass media is leading to noticeable changes in the functioning of contemporary states and societies. Currently, modern technologies play more and more important roles and filter down to almost every field of contemporary human life. It results in the growth of online interactions that can be observed by the inconceivable increase in the number of people with home PCs and Internet access. The proof of it is undoubtedly the emergence and use of concepts such as e-society, e-banking, e-services, e-government, e-government, e-participation and e-democracy. The newly coined word e-democracy evidences that modern technologies have also been widely used in politics. Without any doubt in most countries all actors of political market (politicians, political parties, servants in political/public sector, media) use modern forms of communication with the society. Most of these modern technologies progress the processes of getting and sending information to the citizens, communication with the electorate, and also – which seems to be the biggest advantage – electoral procedures. Thanks to implementation of ICT the interaction between politicians and electorate are improved. The main goal of this text is to analyze electronic voting (e-voting) as one of the important forms of electronic democracy in terms of security aspects. The author of this paper aimed at answering the questions of security of electronic voting as an additional form of participation in elections and referenda.

Keywords: electronic democracy, electronic voting, security of e-voting, information and communication technology (ICT)

Procedia PDF Downloads 203

Authors: Muhammad Luthfi, Sutikno Sutikno, Purhadi Purhadi

Abstract:

Weather forecast has necessarily been improved to provide the communities an accurate and objective prediction as well. To overcome such issue, the numerical-based weather forecast was extensively developed to reduce the subjectivity of forecast. Yet the Numerical Weather Predictions (NWPs) outputs are unfortunately issued without taking dynamical weather behavior and local terrain features into account. Thus, NWPs outputs are not able to accurately forecast the weather quantities, particularly for medium and long range forecast. The aim of this research is to aid and extend the development of ensemble forecast for Meteorology, Climatology, and Geophysics Agency of Indonesia. Ensemble method is an approach combining various deterministic forecast to produce more reliable one. However, such forecast is biased and uncalibrated due to its underdispersive or overdispersive nature. As one of the parametric methods, Bayesian Model Averaging (BMA) generates the calibrated ensemble forecast and constructs predictive PDF for specified period. Such method is able to utilize ensemble of any size but does not take spatial correlation into account. Whereas space dependencies involve the site of interest and nearby site, influenced by dynamic weather behavior. Meanwhile, Geostatistical Output Perturbation (GOP) reckons the spatial correlation to generate future weather quantities, though merely built by a single deterministic forecast, and is able to generate an ensemble of any size as well. This research conducts both BMA and GOP to generate the calibrated ensemble forecast for the daily temperature at few meteorological sites nearby Indonesia international airport.

Keywords: Bayesian Model Averaging, ensemble forecast, geostatistical output perturbation, numerical weather prediction, temperature

Procedia PDF Downloads 248

Authors: Gajendra Kumar Mourya, Dinesh Bhatia, Akash Handique, Sunita Warjri, Syed Achaab Amir

Abstract:

Nowadays, Medical imaging has become an integral part of modern healthcare. Abdominal CT images are an invaluable mean for abdominal organ investigation and have been widely studied in the recent years. Diagnosis of liver pathologies is one of the major areas of current interests in the field of medical image processing and is still an open problem. To deeply study and diagnose the liver, segmentation of liver is done to identify which part of the liver is mostly affected. Manual segmentation of the liver in CT images is time-consuming and suffers from inter- and intra-observer differences. However, automatic or semi-automatic computer aided segmentation of the Liver is a challenging task due to inter-patient Liver shape and size variability. In this paper, we present a technique for automatic segmenting the liver from CT images using Random Forest Classifier. Random forests or random decision forests are an ensemble learning method for classification that operate by constructing a multitude of decision trees at training time and outputting the class that is the mode of the classes of the individual trees. After comparing with various other techniques, it was found that Random Forest Classifier provide a better segmentation results with respect to accuracy and speed. We have done the validation of our results using various techniques and it shows above 89% accuracy in all the cases.

Keywords: CT images, image validation, random forest, segmentation

Procedia PDF Downloads 281

Authors: Isaac Kofi Mensah

Abstract:

The adoption and application of Information and Communication Technologies (ICTs) in government administration through e-government is expected to permeate all sectors of state/ public institutions as well as democratic institutions. One of such public institutions is the Electoral Commission of Ghana mandated by the 1992 Constitution to hold all public elections including presidential and parliamentary elections. As Ghana holds its 7th General Elections since 1992, on 7th November 2016, there are demands from key stakeholders for the Election Management Body, which is the Electoral Commission (EC) of Ghana to adopt and implement an electronic voting system. This case study, therefore, attempts to contribute significantly to the debate by examining influencing factors that would impact on citizen’s readiness to adopt and use an electronic voting system in Ghana. The Technology Acceptance Model (TAM) was used as a theoretical framework for this study, out of which a research model and hypotheses were developed. Importantly, the outcome of this research finding would form a basis for appropriate policy recommendation for consideration of Government and EC of Ghana.

Keywords: citizens readiness, e-government, electronic voting, technology acceptance model (TAM)

Procedia PDF Downloads 229

Authors: Fatma Susilawati Mohamad, Azizah Abdul Manaf, Fadhillah Ahmad, Zarina Mohamad, Wan Suryani Wan Awang

Abstract:

A simple and robust multi-class linear classifier is proposed and implemented. For a pair of classes of the linear boundary, a collection of segments of hyper planes created as perpendicular bisectors of line segments linking centroids of the classes or part of classes. Nearest Neighbor and Linear Discriminant Analysis are compared in the experiments to see the performances of each classifier in discriminating ripeness of oil palm. This paper proposes a multi-class linear classifier using Linear Discriminant Analysis (LDA) for image identification. Result proves that LDA is well capable in separating multi-class features for ripeness identification.

Keywords: multi-class, linear classifier, nearest neighbor, linear discriminant analysis

Procedia PDF Downloads 501

Authors: T. Hari Prasath, P. Ithaya Rani

Abstract:

In the world of visual technology, recognizing emotions from the face images is a challenging task. Several related methods have not utilized the dynamic facial features effectively for high performance. This paper proposes a method for emotions recognition using dynamic facial features with high performance. Initially, local features are captured by Gabor filter with different scale and orientations in each frame for finding the position and scale of face part from different backgrounds. The Gabor features are sent to the ensemble classifier for detecting Gabor facial features. The region of dynamic features is captured from the Gabor facial features in the consecutive frames which represent the dynamic variations of facial appearances. In each region of dynamic features is normalized using Z-score normalization method which is further encoded into binary pattern features with the help of threshold values. The binary features are passed to Multi-class AdaBoost classifier algorithm with the well-trained database contain happiness, sadness, surprise, fear, anger, disgust, and neutral expressions to classify the discriminative dynamic features for emotions recognition. The developed method is deployed on the Ryerson Multimedia Research Lab and Cohn-Kanade databases and they show significant performance improvement owing to their dynamic features when compared with the existing methods.

Keywords: detecting face, Gabor filter, multi-class AdaBoost classifier, Z-score normalization

Procedia PDF Downloads 246

Authors: Andrey V. Timofeev

Abstract:

This paper introduces an original method for guaranteed estimation of the accuracy of an ensemble of Lipschitz classifiers. The solution was obtained as a finite closed set of alternative hypotheses, which contains an object of classification with a probability of not less than the specified value. Thus, the classification is represented by a set of hypothetical classes. In this case, the smaller the cardinality of the discrete set of hypothetical classes is, the higher is the classification accuracy. Experiments have shown that if the cardinality of the classifiers ensemble is increased then the cardinality of this set of hypothetical classes is reduced. The problem of the guaranteed estimation of the accuracy of an ensemble of Lipschitz classifiers is relevant in the multichannel classification of target events in C-OTDR monitoring systems. Results of suggested approach practical usage to accuracy control in C-OTDR monitoring systems are present.

Keywords: Lipschitz classifiers, confidence set, C-OTDR monitoring, classifiers accuracy, classifiers ensemble

Procedia PDF Downloads 461

Authors: Soha A. Bahanshal, Vaibhav Verdhan, Bayong Kim

Abstract:

Graduation rates at six-year colleges are becoming a more essential indicator for incoming fresh students and for university rankings. Predicting student graduation is extremely beneficial to schools and has a huge potential for targeted intervention. It is important for educational institutions since it enables the development of strategic plans that will assist or improve students' performance in achieving their degrees on time (GOT). A first step and a helping hand in extracting useful information from these data and gaining insights into the prediction of students' progress and performance is offered by machine learning techniques. Data analysis and visualization techniques are applied to understand and interpret the data. The data used for the analysis contains students who have graduated in 6 years in the academic year 2017-2018 for science majors. This analysis can be used to predict the graduation of students in the next academic year. Different Predictive modelings such as logistic regression, decision trees, support vector machines, Random Forest, Naïve Bayes, and KNeighborsClassifier are applied to predict whether a student will graduate. These classifiers were evaluated with k folds of 5. The performance of these classifiers was compared based on accuracy measurement. The results indicated that Ensemble Classifier achieves better accuracy, about 91.12%. This GOT prediction model would hopefully be useful to university administration and academics in developing measures for assisting and boosting students' academic performance and ensuring they graduate on time.

Keywords: prediction, decision trees, machine learning, support vector machine, ensemble model, student graduation, GOT graduate on time

Procedia PDF Downloads 48

Authors: Yasser F. Hassan

Abstract:

Social networking involves use of an online platform or website that enables people to communicate, usually for a social purpose, through a variety of services, most of which are web-based and offer opportunities for people to interact over the internet, e.g. via e-mail and ‘instant messaging’, by analyzing the voting behavior and ratings of judges in a popular comments in social networks. While most of the party literature omits the electorate, this paper presents a model where elites and parties are emergent consequences of the behavior and preferences of voters. The research in artificial intelligence and psychology has provided powerful illustrations of the way in which the emergence of intelligent behavior depends on the development of representational structure. As opposed to the classical voting system (one person – one decision – one vote) a new voting system is designed where agents with opposed preferences are endowed with a given number of votes to freely distribute them among some issues. The paper uses ideas from machine learning, artificial intelligence and soft computing to provide a model of the development of voting system response in a simulated agent. The modeled development process involves (simulated) processes of evolution, learning and representation development. The main value of the model is that it provides an illustration of how simple learning processes may lead to the formation of structure. We employ agent-based computer simulation to demonstrate the formation and interaction of coalitions that arise from individual voter preferences. We are interested in coordinating the local behavior of individual agents to provide an appropriate system-level behavior.

Keywords: voting system, rough sets, multi-agent, social networks, emergence, power indices

Procedia PDF Downloads 369

Authors: Priyanka Sharma

Abstract:

The women voter turnout in India is increasing. The rising female voter turnout is explained in part by men intimidating women in the household to vote. Women are more likely than men to be guided before voting. What is perhaps more significant is that the gender gap has shrunk significantly over the years. However, there are layers and categories of women voters in India. Some women are much more likely than the average woman to follow advice. Against this backdrop, this paper investigates the variation among women voters during the national elections of 2019 in India. The central question of this research paper is whether or not the development of greater political opinion among women would offset guided voting and allow them to emerge as more independent voters. So the independent variable of the study is Indian women’s opinion on politics, and the dependent variable is their voting behavior. The methodology used in this paper is both quantitative and qualitative. This study investigated and examined Lokniti’s election survey data. The sample size used in this survey is 11568. The analysis of this study has revealed that there is a considerable impact of women having a political opinion on their voting behavior. The Bivariate analysis of the variables states that 83% of Indian women who have opinions on political issues do not seek advice while going to vote. This proves the hypothesis of this paper that women with an opinion on politics are more likely to be independent voters. To check the statistical significance of the finding, a chi-square test was done and the p-value found is 0.009737, which shows it is statistically significant. Furthermore, a regression test has been done by controlling certain variables like age, educational qualification, caste, and financial position of the women to probe the influence on the dependent variable. The findings provide worthwhile insights into the relationship between these control variables and the women voting behavior in India.

Keywords: dependent voter, independent voter, political opinion, voting behavior, women voter

Procedia PDF Downloads 55

Authors: Myungjin Lee, Daegun Han, Jongsung Kim, Soojun Kim, Hung Soo Kim

Abstract:

Recently, the localized heavy rainfall and typhoons are frequently occurred due to the climate change and the damage is becoming bigger. Therefore, we may need a more accurate prediction of the rainfall and runoff. However, the gauge rainfall has the limited accuracy in space. Radar rainfall is better than gauge rainfall for the explanation of the spatial variability of rainfall but it is mostly underestimated with the uncertainty involved. Therefore, the ensemble of radar rainfall was simulated using error structure to overcome the uncertainty and gauge rainfall. The simulated ensemble was used as the input data of the rainfall-runoff models for obtaining the ensemble of runoff hydrographs. The previous studies discussed about the accuracy of the rainfall-runoff model. Even if the same input data such as rainfall is used for the runoff analysis using the models in the same basin, the models can have different results because of the uncertainty involved in the models. Therefore, we used two models of the SSARR model which is the lumped model, and the Vflo model which is a distributed model and tried to simulate the optimum runoff considering the uncertainty of each rainfall-runoff model. The study basin is located in Han river basin and we obtained one integrated runoff hydrograph which is an optimum runoff hydrograph using the blending methods such as Multi-Model Super Ensemble (MMSE), Simple Model Average (SMA), Mean Square Error (MSE). From this study, we could confirm the accuracy of rainfall and rainfall-runoff model using ensemble scenario and various rainfall-runoff model and we can use this result to study flood control measure due to climate change. Acknowledgements: This work is supported by the Korea Agency for Infrastructure Technology Advancement(KAIA) grant funded by the Ministry of Land, Infrastructure and Transport (Grant 18AWMP-B083066-05).

Keywords: radar rainfall ensemble, rainfall-runoff models, blending method, optimum runoff hydrograph

Procedia PDF Downloads 245

Authors: Chien-Kuo Chang, Yu-Hsiang Lin, Yi-Yun Tang, Min-Chiu Wu

Abstract:

With the trend of technological advancement, the harm caused by power outages is substantial, mostly due to problems in the power grid. This highlights the necessity for further improvement in the reliability of the power system. In the power system, gas-insulated switches (GIS) and power cables play a crucial role. Long-term operation under high voltage can cause insulation materials in the equipment to crack, potentially leading to partial discharges. If these partial discharges (PD) can be analyzed, preventative maintenance and replacement of equipment can be carried out, there by improving the reliability of the power grid. This research will diagnose defects by identifying three different defects in GIS and three different defects in straight cable joints, for a total of six types of defects. The partial discharge data measured will be converted through phase analysis diagrams and pulse sequence analysis. Discharge features will be extracted using convolutional image processing, and three different deep learning models, CNN, ResNet18, and MobileNet, will be used for training and evaluation. Class Activation Mapping will be utilized to interpret the black-box problem of deep learning models, with each model achieving an accuracy rate of over 95%. Lastly, the overall model performance will be enhanced through an ensemble learning voting method.

Keywords: partial discharge, gas-insulated switches, straight cable joint, defect identification, deep learning, ensemble learning

Procedia PDF Downloads 42

Authors: Felix O. Omosele

Abstract:

Several studies have established that electronic voting machines are prone to multi-faceted challenges. One of which is their capacity to lose votes after the ballots might have been cast. Therefore, the international consensus appears to favour the use of electronic voting machines that are accompanied with verifiable audit paper audit trail (VVPAT). At present, there is no known study that has evaluated the impacts (or otherwise) of this verification and auditing on the authentication, admissibility and evidential weight of electronically-obtained electoral data. This legal inquiry is important as elections are sometimes won or lost in courts and on the basis of such data. This gap will be filled by the present research work. Using the United States of America as a case study, this paper employed a qualitative legal analysis of several of its appellate courts’ judicial opinions. This analysis equally unearths the necessary statutory rules and regulations that are important to the research problem. The objective of the research is to highlight the roles played by VVPAT on electoral evidence- as seen from the eyes of the court. The preliminary outcome of this qualitative analysis shows that the admissibility and weight attached to ‘Computer Evidence from e-voting machines (CEEM)’ are often treated with general standards applied to other computer-stored evidence. These standards sometimes fail to embrace the peculiar challenges faced by CEEM, particularly with respect to their tabulation and transmission. This paper, therefore, argues that CEEM should be accorded unique consideration by courts. It proposes the development of a legal standard which recognises verification and auditing as ‘weight enhancers’ for electronically-obtained electoral data.

Keywords: admissibility of computer evidence, electronic voting, qualitative legal analysis, voting machines in the USA

Procedia PDF Downloads 166

Authors: Rodrigo S. Moreira, Nelson F. F. Ebecken

Abstract:

This paper proposes a maritime vessel tracker composed of an ensemble of WiSARD weightless neural network classifiers. A failure detector analyzes vessel movement with a Kalman filter and corrects the tracking, if necessary, using FFT matching. The use of the WiSARD neural network to track objects is uncommon. The additional contributions of the present study include a performance comparison with four state-of-art trackers, an experimental study of the features that improve maritime vessel tracking, the first use of an ensemble of classifiers to track maritime vessels and a new quantization algorithm that compares the values of pixel pairs.

Keywords: ram memory, WiSARD weightless neural network, object tracking, quantization

Procedia PDF Downloads 280

Authors: Shaukat Ali Shahee, Usha Ananthakumar

Abstract:

A data set exhibits class imbalance problem when one class has very few examples compared to the other class, and this is also referred to as between class imbalance. The traditional classifiers fail to classify the minority class examples correctly due to its bias towards the majority class. Apart from between-class imbalance, imbalance within classes where classes are composed of a different number of sub-clusters with these sub-clusters containing different number of examples also deteriorates the performance of the classifier. Previously, many methods have been proposed for handling imbalanced dataset problem. These methods can be classified into four categories: data preprocessing, algorithmic based, cost-based methods and ensemble of classifier. Data preprocessing techniques have shown great potential as they attempt to improve data distribution rather than the classifier. Data preprocessing technique handles class imbalance either by increasing the minority class examples or by decreasing the majority class examples. Decreasing the majority class examples lead to loss of information and also when minority class has an absolute rarity, removing the majority class examples is generally not recommended. Existing methods available for handling class imbalance do not address both between-class imbalance and within-class imbalance simultaneously. In this paper, we propose a method that handles between class imbalance and within class imbalance simultaneously for binary classification problem. Removing between class imbalance and within class imbalance simultaneously eliminates the biases of the classifier towards bigger sub-clusters by minimizing the error domination of bigger sub-clusters in total error. The proposed method uses model-based clustering to find the presence of sub-clusters or sub-concepts in the dataset. The number of examples oversampled among the sub-clusters is determined based on the complexity of sub-clusters. The method also takes into consideration the scatter of the data in the feature space and also adaptively copes up with unseen test data using Lowner-John ellipsoid for increasing the accuracy of the classifier. In this study, neural network is being used as this is one such classifier where the total error is minimized and removing the between-class imbalance and within class imbalance simultaneously help the classifier in giving equal weight to all the sub-clusters irrespective of the classes. The proposed method is validated on 9 publicly available data sets and compared with three existing oversampling techniques that rely on the spatial location of minority class examples in the euclidean feature space. The experimental results show the proposed method to be statistically significantly superior to other methods in terms of various accuracy measures. Thus the proposed method can serve as a good alternative to handle various problem domains like credit scoring, customer churn prediction, financial distress, etc., that typically involve imbalanced data sets.

Keywords: classification, imbalanced dataset, Lowner-John ellipsoid, model based clustering, oversampling

Procedia PDF Downloads 389

Authors: Amin Salighehdar, Ziwen Ye, Mingzhe Liu, Ionut Florescu, Alan F. Blumberg

Abstract:

Storm surge is an abnormal water level caused by a storm. Accurate prediction of a storm surge is a challenging problem. Researchers developed various ensemble modeling techniques to combine several individual forecasts to produce an overall presumably better forecast. There exist some simple ensemble modeling techniques in literature. For instance, Model Output Statistics (MOS), and running mean-bias removal are widely used techniques in storm surge prediction domain. However, these methods have some drawbacks. For instance, MOS is based on multiple linear regression and it needs a long period of training data. To overcome the shortcomings of these simple methods, researchers propose some advanced methods. For instance, ENSURF (Ensemble SURge Forecast) is a multi-model application for sea level forecast. This application creates a better forecast of sea level using a combination of several instances of the Bayesian Model Averaging (BMA). An ensemble dressing method is based on identifying best member forecast and using it for prediction. Our contribution in this paper can be summarized as follows. First, we investigate whether the ensemble models perform better than any single forecast. Therefore, we need to identify the single best forecast. We present a methodology based on a simple Bayesian selection method to select the best single forecast. Second, we present several new and simple ways to construct ensemble models. We use correlation and standard deviation as weights in combining different forecast models. Third, we use these ensembles and compare with several existing models in literature to forecast storm surge level. We then investigate whether developing a complex ensemble model is indeed needed. To achieve this goal, we use a simple average (one of the simplest and widely used ensemble model) as benchmark. Predicting the peak level of Surge during a storm as well as the precise time at which this peak level takes place is crucial, thus we develop a statistical platform to compare the performance of various ensemble methods. This statistical analysis is based on root mean square error of the ensemble forecast during the testing period and on the magnitude and timing of the forecasted peak surge compared to the actual time and peak. In this work, we analyze four hurricanes: hurricanes Irene and Lee in 2011, hurricane Sandy in 2012, and hurricane Joaquin in 2015. Since hurricane Irene developed at the end of August 2011 and hurricane Lee started just after Irene at the beginning of September 2011, in this study we consider them as a single contiguous hurricane event. The data set used for this study is generated by the New York Harbor Observing and Prediction System (NYHOPS). We find that even the simplest possible way of creating an ensemble produces results superior to any single forecast. We also show that the ensemble models we propose generally have better performance compared to the simple average ensemble technique.

Keywords: Bayesian learning, ensemble model, statistical analysis, storm surge prediction

Procedia PDF Downloads 285

Authors: Arian Hosseini, Mahmudul Hasan

Abstract:

To address the increasing need for efficient and accurate content moderation, we propose an efficient and lightweight deep classification ensemble structure. Our approach is based on a combination of simple visual features, designed for high-accuracy classification of violent content with low false positives. Our ensemble architecture utilizes a set of lightweight models with narrowed-down color features, and we apply it to both images and videos. We evaluated our approach using a large dataset of explosion and blast contents and compared its performance to popular deep learning models such as ResNet-50. Our evaluation results demonstrate significant improvements in prediction accuracy, while benefiting from 7.64x faster inference and lower computation cost. While our approach is tailored to explosion detection, it can be applied to other similar content moderation and violence detection use cases as well. Based on our experiments, we propose a "think small, think many" philosophy in classification scenarios. We argue that transforming a single, large, monolithic deep model into a verification-based step model ensemble of multiple small, simple, and lightweight models with narrowed-down visual features can possibly lead to predictions with higher accuracy.

Keywords: deep classification, content moderation, ensemble learning, explosion detection, video processing

Procedia PDF Downloads 21

Authors: Nidal F. Shilbayeh, Reem A. Al-Saidi, Ahmed H. Alsswey

Abstract:

In this paper, we presented an evaluation and analysis of E-Voting Authentication Preparation Scheme (EV-APS). EV-APS applies some modified security aspects that enhance the security measures and adds a strong wall of protection, confidentiality, non-repudiation and authentication requirements. Some of these modified security aspects are Kerberos authentication protocol, PVID scheme, responder certificate validation, and the converted Ferguson e-cash protocol. Authentication and privacy requirements have been evaluated and proved. Authentication guaranteed only eligible and authorized voters were permitted to vote. Also, the privacy guaranteed that all votes will be kept secret. Evaluation and analysis of some of these security requirements have been given. These modified aspects will help in filtering the counter buffer from unauthorized votes by ensuring that only authorized voters are permitted to vote.

Keywords: e-voting preparation stage, blind signature protocol, Nonce based authentication scheme, Kerberos Authentication Protocol, pseudo voter identity scheme PVID

Procedia PDF Downloads 267

Authors: Uswah Khairuddin, Rubiyah Yusof, Nenny Ruthfalydia Rosli

Abstract:

This paper presents a comparative study between k-Nearest Neighbour (k-NN) and Multi-Layer Perceptron Neural Network (MLP-NN) classifier using Genetic Algorithm (GA) as feature selector for wood recognition system. The features have been extracted from the images using Grey Level Co-Occurrence Matrix (GLCM). The use of GA based feature selection is mainly to ensure that the database used for training the features for the wood species pattern classifier consists of only optimized features. The feature selection process is aimed at selecting only the most discriminating features of the wood species to reduce the confusion for the pattern classifier. This feature selection approach maintains the ‘good’ features that minimizes the inter-class distance and maximizes the intra-class distance. Wrapper GA is used with k-NN classifier as fitness evaluator (GA-kNN). The results shows that k-NN is the best choice of classifier because it uses a very simple distance calculation algorithm and classification tasks can be done in a short time with good classification accuracy.

Keywords: feature selection, genetic algorithm, optimization, wood recognition system

Procedia PDF Downloads 509

Authors: G. Singer, M. Golan

Abstract:

Providing students with learning disabilities (LD) with extra time to grant them equal access to the exam is a necessary but insufficient condition to compensate for their LD; there should also be a clear indication that the additional time was actually used. For example, if students with LD use more time than students without LD and yet receive lower grades, this may indicate that a different accommodation is required. If they achieve higher grades but use the same amount of time, then the effectiveness of the accommodation has not been demonstrated. The main goal of this study is to evaluate the effect of including parameters related to LD and extended exam time, along with other commonly-used characteristics (e.g., student background and ability measures such as high-school grades), on the ability of ordinal decision-tree algorithms to predict exam performance. We use naturally-occurring data collected from hundreds of undergraduate engineering students. The sub-goals are i) to examine the improvement in prediction accuracy when the indicator of exam performance includes 'actual time used' in addition to the conventional indicator (exam grade) employed in most research; ii) to explore the effectiveness of extended exam time on exam performance for different courses and for LD students with different profiles (i.e., sets of characteristics). This is achieved by using the patterns (i.e., subgroups) generated by the algorithms to identify pairs of subgroups that differ in just one characteristic (e.g., course or type of LD) but have different outcomes in terms of exam performance (grade and time used). Since grade and time used to exhibit an ordering form, we propose a method based on ordinal decision-trees, which applies a weighted information-gain ratio (WIGR) measure for selecting the classifying attributes. Unlike other known ordinal algorithms, our method does not assume monotonicity in the data. The proposed WIGR is an extension of an information-theoretic measure, in the sense that it adjusts to the case of an ordinal target and takes into account the error severity between two different target classes. Specifically, we use ordinal C4.5, random-forest, and AdaBoost algorithms, as well as an ensemble technique composed of ordinal and non-ordinal classifiers. Firstly, we find that the inclusion of LD and extended exam-time parameters improves prediction of exam performance (compared to specifications of the algorithms that do not include these variables). Secondly, when the indicator of exam performance includes 'actual time used' together with grade (as opposed to grade only), the prediction accuracy improves. Thirdly, our subgroup analyses show clear differences in the effect of extended exam time on exam performance among different courses and different student profiles. From a methodological perspective, we find that the ordinal decision-tree based algorithms outperform their conventional, non-ordinal counterparts. Further, we demonstrate that the ensemble-based approach leverages the strengths of each type of classifier (ordinal and non-ordinal) and yields better performance than each classifier individually.

Keywords: actual exam time usage, ensemble learning, learning disabilities, ordinal classification, time extension

Procedia PDF Downloads 68

Authors: Jason Sah

Abstract:

Music ensemble class is a non-traditional classroom in the sense that it is always a group effort during rehearsal. When measuring student performance ability in class, it is imperative that the grading rubric includes a collaborative skill component. Assessments that stop short of testing students' ability to make music with others undermine the group mentality by elevating individual prowess. Applying empirical and evidence-based methodology, this research develops a grading rubric that defines the criteria for assessing collaborative skill, and then explores different strategies for implementing this rubric in a timely and effective manner. Findings show that when collaborative skill is regularly tested, students gradually shift their attention from playing their own part well to sharing their part with others.

Keywords: assessment, ensemble class, grading rubric, student collaboration

Procedia PDF Downloads 107

Authors: Prateek Bhattacharya

Abstract:

The definition of 'control' in India is a rapidly evolving concept, owing to varying rights attached to varying securities. Shares with differential voting rights (DVRs) provide the holder with differential rights as to voting, as compared to ordinary equity shareholders of the company. Such DVRs can amount to both superior voting rights and inferior voting rights, where DVRs with superior voting rights amount to providing the holder with golden shares in the company. While DVRs are not a novel concept in India having been recognized since 2000, they were placed on a back burner by the Securities and Exchange Board of India (SEBI) in 2010 after issuance of DVRs with superior voting rights was restricted. In June 2019, the SEBI rekindled the ebbing fire of DVRs, keeping mind the fast-paced nature of the global economy, the government's faith that India’s ‘new age technology companies’ (i.e., Start-Ups) will lead the charge in achieving its goal of India becoming a $5 trillion dollar economy by 2024, and recognizing that the promoters of such Start-Ups seek to raise capital without losing control over their companies. DVRs with superior voting rights guarantee promoters with up to 74% shareholding in Start-Ups for a period of 5 years, meaning that the holder of such DVRs can exercise sole control and material influence over the company for that period. This manner of control has the potential of causing both pro-competitive and anti-competitive effects in the markets where these companies operate. On the one hand, DVRs will allow Start-Up promoters/founders to retain control of their companies and protect its business interests from foreign elements such as private/public investors – in a scenario where such investors have multiple investments in firms engaged in associated lines of business (whether on a horizontal or vertical level) and would seek to influence these firms to enter into potential anti-competitive arrangements with one another, DVRs will enable the promoters to thwart such scenarios. On the other hand, promoters/founders who themselves have multiple investments in Start-Ups, which are in associated lines of business run the risk of influencing these associated Start-Ups to engage in potentially anti-competitive arrangements in the name of profit maximisation. This paper shall be divided into three parts: Part I shall deal with the concept of ‘control’, as deliberated upon and decided by the SEBI and the Competition Commission of India (CCI) under both company/securities law and competition law; Part II shall review this definition of ‘control’ through the lens of DVRs, and Part III shall discuss the aforementioned potential pro-competitive and anti-competitive effects caused by the DVRs by examining the current Indian Start-Up scenario. The paper shall conclude by providing suggestions for the CCI to incorporate a clearer and more progressive concept of ‘control’.

Keywords: competition law, competitive effects, control, differential voting rights, DVRs, investor shareholding, merger control, start-ups

Procedia PDF Downloads 96

Authors: Saeedeh Barzegar Khalilsaraei, Manoocheher Kelarestaghi, Farshad Eshghi

Abstract:

The vehicle color, as a prominent and stable feature, helps to identify a vehicle more accurately. As a result, vehicle color recognition is of great importance in intelligent transportation systems. Unlike conventional methods which use only a single Convolutional Neural Network (CNN) for feature extraction or classification, in this paper, four CNNs, with different architectures well-performing in different classes, are trained to extract various features from the input image. To take advantage of the distinct capability of each network, the multiple outputs are combined using a stack generalization algorithm as an ensemble technique. As a result, the final model performs better than each CNN individually in vehicle color identification. The evaluation results in terms of overall average accuracy and accuracy variance show the proposed method’s outperformance compared to the state-of-the-art rivals.

Keywords: Vehicle Color Recognition, Ensemble Algorithm, Stack Generalization, Convolutional Neural Network

Procedia PDF Downloads 51

Authors: Atefeh Heydari, Mohammadali Tavakoli, Zuriati Ismail, Naomie Salim

Abstract:

Seeking and sharing knowledge on online forums have made them popular in recent years. Although online forums are valuable sources of information, due to variety of sources of messages, retrieving reliable threads with high quality content is an issue. Majority of the existing information retrieval systems ignore the quality of retrieved documents, particularly, in the field of thread retrieval. In this research, we present an approach that employs various quality features in order to investigate the quality of retrieved threads. Different aspects of content quality, including completeness, comprehensiveness, and politeness, are assessed using these features, which lead to finding not only textual, but also conceptual relevant threads for a user query within a forum. To analyse the influence of the features, we used an adopted version of voting model thread search as a retrieval system. We equipped it with each feature solely and also various combinations of features in turn during multiple runs. The results show that incorporating the quality features enhances the effectiveness of the utilised retrieval system significantly.

Keywords: content quality, forum search, thread retrieval, voting techniques

Procedia PDF Downloads 181

Authors: Yash Jain

Abstract:

The US Center for Disease Control has recommended wearing masks to slow the spread of the virus. The research uses a video feed from a camera to conduct real-time classifications of whether or not a human is correctly wearing a mask, incorrectly wearing a mask, or not wearing a mask at all. Utilizing two distinct datasets from the open-source website Kaggle, a mask detection network had been trained. The first dataset that was used to train the model was titled 'Face Mask Detection' on Kaggle, where the dataset was retrieved from and the second dataset was titled 'Face Mask Dataset, which provided the data in a (YOLO Format)' so that the TinyYoloV3 model could be trained. Based on the data from Kaggle, two machine learning models were implemented and trained: a Tiny YoloV3 Real-time model and a two-stage neural network classifier. The two-stage neural network classifier had a first step of identifying distinct faces within the image, and the second step was a classifier to detect the state of the mask on the face and whether it was worn correctly, incorrectly, or no mask at all. The TinyYoloV3 was used for the live feed as well as for a comparison standpoint against the previous two-stage classifier and was trained using the darknet neural network framework. The two-stage classifier attained a mean average precision (MAP) of 80%, while the model trained using TinyYoloV3 real-time detection had a mean average precision (MAP) of 59%. Overall, both models were able to correctly classify stages/scenarios of no mask, mask, and incorrectly worn masks.

Keywords: datasets, classifier, mask-detection, real-time, TinyYoloV3, two-stage neural network classifier

Procedia PDF Downloads 127