Performance Of Classification Algorithms Research Articles

Video Shot Boundary Detection and Sports Video Classification Algorithm Based on Particle Filter

INTRODUCTION: Sports video is a very important information resource. The classification of sports video with high accuracy can effectively improve the browsing and query effect of users. This project intends to study a motion video classification algorithm based on deep learning particle filter to solve the problems of strong subjectivity and low accuracy of existing motion video classification algorithms. A key box extraction method based on similarity is proposed. The moving video classification algorithm based on deep learning coding model is studied. Examples of various types of sports videos are analyzed. The overall performance of the motion video classification algorithm proposed in this paper is much better than other existing motion video classification algorithms. This algorithm can greatly improve the classification performance of motion video.

Research on the diagnosis of corona virus disease 2019(COVID-19) based on machine-learning

In recent years, the Corona Virus Disease 2019(COVID-19) has brought a huge impact on people's daily life and the normal operation of society, and as machine learning research deepens, this technology could help detect viruses such as the novel coronavirus. According to this, how to accurately, quickly and effectively analyze and classify laboratory results with multiple indicators through the method of machine learning is the object of this paper. In order to explore the classification performance of various machine learning classification algorithms on laboratory results, linear and non-linear methods were used respectively to analyze and classify the laboratory results. In linear analysis, distance discriminant and fisher discriminant were used to explore the classification effect of linear classification on laboratory results. The non-linear analysis mainly used Adaptive Boosting(AdaBoost) and Random Forest algorithm which are widely used to test the classification effect under the influence of multiple indexes. In this paper, python and other tools were used to classify samples of different combinations by using the idea of cross validation. By comparing the running time and detection accuracy, it was found that Adaboost algorithm is applicable in most cases and was a relatively fast and accurate classification method. In addition, Random Forest algorithm had similar accuracy, but it might have better performance on a large data set.

A BiLSTM-CF and BiGRU-based Deep Sentiment Analysis Model to Explore Customer Reviews for Effective Recommendations

The advancement of technology has led to the rise of social media forums and e-commerce platforms, which have become popular means of communication, and people can express their opinions through comments and reviews. Increased accessibility to online feedback helps individuals make informed decisions about product purchases, services, and other decisions. This study used a sentiment analysis-based approach to improve the functionality of the recommendations from user reviews and consider the features (aspects and opinions) of products and services to understand the characteristics and attributes that influence the performance of classification algorithms. The proposed model consists of data preprocessing, word embedding, character representation creation, feature extraction using BiLSTM-CF, and classification using BiGRU. The proposed model was evaluated on different multidomain benchmark datasets demonstrating impressive performance. The proposed model outperformed existing models, offering more promising performance results in recommendations.

Generative modelling of vibration signals in machine maintenance

The exponential development of technologies for the acquisition, collection, and processing of data from real-world objects is creating new perspectives in the field of machine maintenance. The Industrial Internet of Things is the source of a huge collection of measurement data. The performance of classification or regression algorithms needs to take into account the random nature of the process being modelled and any incomplete observability, especially in terms of failure states. The article highlights the practical possibilities of using generative artificial intelligence and deep machine learning systems to create synthetic measurement observations in monitoring the vibrations of rotating machinery to improve unbalanced databases. Variational Autoencoder VAE generative models with latent variables in the form of high-level input features of time-frequency spectra were studied. The mapping and generation algorithm was optimised and its effectiveness was tested in the practical solution of the task of diagnosing the three operating states of a demonstration gearbox.

Detection of influential observations in high-dimensional survival data

Survival analysis is a statistical technique mainly used to analyze time-to-event data. Identification of influential observation attains greater importance since it leads to discovering new prognostic factors. Influential observation in survival typically points to individuals whose survival time is extremely short or long in comparison to others. Particularly, when the data possess more covariates than the observations, all classical approaches fail to perform. Hence, dimensionality reduction is necessary for choosing appropriate variables and it has been done by popular techniques such as LASSO and elastic net algorithm. This paper consider high-dimensional breast cancer data, and its dimensionality is reduced using variable selection methods. Subsequently, the rank product test and martingale residuals are used to identify an influential observation. Furthermore, a resampling technique is used to validate the consistency and robustness of the methods. The novelty of this paper lies in comparing the prediction accuracy of datasets with and without outliers using Random Survival Forest (RSF) for different training fractions. Comparatively, the RSF result demonstrates that the LASSO approach outperform others in the absence of outliers. Therefore, we suggest reducing dimensionality using the LASSO variable selection technique first, followed by removing likely outliers to improve the performance of classification algorithms.

Comparison of Algorithms for Sentiment Analysis of Operator Satisfaction Level for Increasing Neo Feeder Applications in PDDikti Higher Education LLDIKTI Region VI Semarang Central Java

Sentiment analysis on the satisfaction level of PDDikti operators is very important to find out how PDDikti operators feel after the version of the academic reporting application for higher education was upgraded, namely Neo Feeeder. The increase in the version of this application causes some of the features in it to not function properly. So some academic reporting activities from tertiary institutions experience problems. As a result of this condition, the most felt impact is students, where students experience delays in graduation. Then it is necessary to evaluate through sentiment analysis from PDDikti operators to find out the response from operators and be able to provide positive suggestions to developers from the PDDikti reporting application. This study applies several classification methods for sentiment analysis at once, including the Random Forest algorithm, the Support Vector Machine algorithm, the Multinomial Naïve Bayes algorithm, the Decision Tree algorithm, and the K-Nearest Neighbor algorithm. Of the 5 methods applied, the results of their performance accuracy will be compared. The performance of the highest classification algorithm is the K-Nearest Neighbor (K-NN) algorithm which produces an accuracy value when testing data, which is up to 90% using the oversampling technique in unbalanced classes. While the lowest classification accuracy performance value is in the Multinomial Naïve Bayes (MNB) algorithm with a value of 76%. It is proven that oversampling can help the performance of the classification algorithm to be more optimal. Thus, it should be noted that the balance of data classes is an important factor when applying the classification method.

Optimized Feature Selection for PCOS Disease Prediction

PCOS is the most predominant endocrine problem among women of reproductive age, which affects ovaries and causes irregularities in menstrual cycles, weight gain, and hirsutism [Bulsara J., P. Patel, A. Soni, S. Acharya (2021) A review: Brief insight into PCOS, Endocrine and Metabolic Sci., 3, 1–7]. The condition probably results from a mix of causes, including qualities and ecological elements. The aim of this study is to investigate the performance of Binarized Butterfly Optimization Algorithm, Binarized Grey Wolf Optimization Algorithm, Binarized Genetic Algorithm and Binarized Cuckoo Optimization Algorithm in terms of classification accuracy. Kaggle PCOS dataset is used for this work and it has 541 records and 43 attributes. The study aims to investigate the performance of four optimization algorithms in terms of classification accuracy for predicting PCOS disease. The BGWO optimization algorithm outperformed the other algorithms with 99% accuracy, while the BGA produced the most optimal subset. The study highlights the significance of feature selection in improving the performance of classification algorithms.

Development of Machine Learning Algorithms for Riverside Land Cover Classification Using Synthetic Aperture Radar Satellite Imagery and Terrain Data

Riverine environments play a crucial role in maintaining the stability of river ecosystems as well as biodiversity. Furthermore, the appropriate management of small rivers has a significant impact not only on stable water supplies but also on water resource management. Wide monitoring of the riverside environment including land covers and their changes is an important issue in water resource management. This study aims to develop a high-resolution (10 m) model for classifying riverside land cover by integrating Sentinel-1 synthetic aperture radar (SAR) data and terrestrial characteristics using machine learning algorithms. We constructed a total of 3,284 landcover reference point datasets near the four major rivers of South Korea with five classes: water, barren, grass, forest, and built-up. The Random Forest and Light Gradient Boosting Machine classification models were developed using eight input variables derived from SAR signal and digital terrain data. The models showed an overall cross-validation accuracy exceeding 80% while maintaining consistent spatial distributions, except for the barren class. The false alarms on barren would be corrected through additional sampling processes and incorporating optical characteristics in further study. The high-resolution riverside land cover maps are expected to contribute to the establishment of a comprehensive management system for water resources such as riverside land cover change detection, river ecosystem monitoring, and flood hazard management. Furthermore, the utilization of the next generation medium satellite 5 (C-band SAR) would improve the performance of riverside land cover classification algorithm in the future.

An Educational Approach to Higgs Boson Hunting Using Machine Learning Classification Algorithms on ATLAS Open Data

In this study, the performance of several classification algorithms that are used to separate the H → ττ signal from background is investigated. The data set came from the publicly available ATLAS data, which was utilized for the Machine Learning (ML) competition. The data was obtained from a full ATLAS simulation and originated from proton-proton collisions. There are 250 thousand events in the data set, and 70% of them were used to train the algorithms. The primary objective of this research is to identify the signal events from the background events by using various ML methods in the context of high-energy physics. In order to discover a solution to the binary classification problem that was discussed earlier, six distinct classification algorithms were utilized. This article also compares the performance of these classification algorithms, including Linear Support Vector Machines (SVM), Radical SVM, Logistic Regression, K-Nearest Neighbours, XGBoost Classifier, and the AdaBoost Classifier. The best results were obtained using the XGBoost Classification method, which had an AUC of 0.84 ± 1.9 x 10-3 followed by the AdaBoost Classifier with an AUC of 0.82 ± 2.5 x 10-3.

Boosting the visibility of services in microservice architecture

Monolithic software architectures are no longer sufficient for the highly complex software-intensive systems, which modern society depends on. Service Oriented Architecture (SOA) surpassed monolithic architecture due to its reusability, platform independency, ease of maintenance, and scalability. Recent SOA implementations made use of cloud-native architectural approaches such as microservice architecture, which has resulted in a new challenge: the discovery difficulties of services. One way to dynamically discover and route traffic to service instances is to use a service discovery tool to locate the Internet Protocol (IP) address and port number of a microservice. In the event that replicated microservice instances are found to provide the same function, it is crucial to select the right microservice that provides the best overall experience for the end-user. Parameters including success rate, efficiency, delay time, and response time play a vital role in establishing a microservice’s Quality of Service (QoS). These assessments can be performed by means of a live health-check service, or, alternatively, by making a prediction of the current state of affairs with the application of machine learning-based approaches. In this research, we evaluate the performance of several classification algorithms for estimating the quality of microservices using the QWS dataset containing traffic data of 2505 microservices. Our research also analyzed the boosting algorithms, namely Gradient Boost, XGBoost, LightGBM, and CatBoost to improve the overall performance. We utilized parameter optimization techniques, namely Grid Search, Random Search, Bayes Search, Halvin Grid Search, and Halvin Random Search to fine-tune the hyperparameters of our classifier models. Experimental results demonstrated that the CatBoost algorithm achieved the highest level of accuracy (90.42%) in predicting microservice quality.

Improving the accuracy of sentiment analysis using a linguistic rule-based feature selection method in tourism reviews

Sentiment Analysis technique involves extracting the relevant information from Unstructured User Reviews (UUR) dataset fetched from online and classifying them into appropriate positive and negative comments for making decisions. In UUR, data may be in noisy state, irrelevant features exist which creates high dimensional feature space. To design an effective sentiment learning model, users are required to extract the most relevant sentiment features from UUR. To overcome the issue, we proposed a Linguistic rule based feature selection method for extracting and selecting the sentiment features for Sentiment Analysis as it improves the predictive performance of classification algorithms. The proposed novel feature selection method involves identifying the various sentiment features in the review dataset by using filtering methods such as POS tags, n-grams. In the ensemble model, where the Random Forest classification algorithm is trained for textual sentiment classification, the chosen sentiment feature sets are used. Finally, we test our approach using the real-time review dataset that was collected from a multitude of sources, and the results demonstrate prediction accuracy that is superior to that of existing Sentiment analysis techniques.

Validation of a Smartphone Pupillometry Application in Diagnosing Severe Traumatic Brain Injury.

The pupillary light reflex (PLR) is an important biomarker for the detection and management of traumatic brain injury (TBI). We investigated the performance of PupilScreen, a smartphone-based pupillometry app, in classifying healthy control subjects and subjects with severe TBI in comparison to the current gold standard NeurOptics pupillometer (NPi-200 model with proprietary Neurological Pupil Index [NPi] TBI severity score). A total of 230 PLR video recordings taken using both the PupilScreen smartphone pupillometer and NeurOptics handheld device (NPi-200) pupillometer were collected from 33 subjects with severe TBI (sTBI) and 132 subjects who were healthy without self-reported neurological disease. Severe TBI status was determined by Glasgow Coma Scale (GCS) at the time of recording. The proprietary NPi score was collected from the NPi-200 pupillometer for each subject. Seven PLR curve morphological parameters were collected from the PupilScreen app for each subject. A comparison via t-test and via binary classification algorithm performance using NPi scores from the NPi-200 and PLR parameter data from the PupilScreen app was completed. This was used to determine how the frequently used NPi-200 proprietary NPi TBI severity score compares to the PupilScreen app in ability to distinguish between healthy and sTBI subjects. Binary classification models for this task were trained for the diagnosis of healthy or severe TBI using logistic regression, k-nearest neighbors, support vector machine, and random forest machine learning classification models. Overall classification accuracy, sensitivity, specificity, area under the curve, and F1 score values were calculated. Median GCS was 15 for the healthy cohort and 6 (interquartile range 2) for the severe TBI cohort. Smartphone app PLR parameters as well as NPi from the digital infrared pupillometer were significantly different between healthy and severe TBI cohorts; 33% of the study cohort had dark eye colors defined as brown eyes of varying shades. Across all classification models, the top performing PLR parameter combination for classifying subjects as healthy or sTBI for PupilScreen was maximum diameter, constriction velocity, maximum constriction velocity, and dilation velocity with accuracy, sensitivity, specificity, area under the curve (AUC), and F1 score of 87%, 85.9%, 88%, 0.869, and 0.85, respectively, in a random forest model. The proprietary NPi TBI severity score demonstrated greatest AUC value, F1 score, and sensitivity of 0.648, 0.567, and 50.9% respectively using a random forest classifier and greatest overall accuracy and specificity of 67.4% and 92.4% using a logistic regression model in the same classification task on the same dataset. The PupilScreen smartphone pupillometry app demonstrated binary healthy versus severe TBI classification ability greater than that of the NPi-200 proprietary NPi TBI severity score. These results may indicate the potential benefit of future study of this PupilScreen smartphone pupillometry application in comparison to the NPi-200 digital infrared pupillometer across the broader TBI spectrum, as well as in other neurological diseases.

Feature Selection Using Firefly Algorithm With Tree-Based Classification In Software Defect Prediction

Defects that occur in software products are a universal occurrence. Software defect prediction is usually carried out to determine the performance, accuracy, precision and performance of the prediction model or method used in research, using various kinds of datasets. Software defect prediction is one of the Software Engineering studies that is of great concern to researchers. This research was conducted to determine the performance of tree-based classification algorithms including Decision Trees, Random Forests and Deep Forests without using feature selection and using firefly feature selection. And also know the tree-based classification algorithm with firefly feature selection which can provide better software defect prediction performance. The dataset used in this study is the ReLink dataset which consists of Apache, Safe and Zxing. Then the data is divided into testing data and training data with 10-fold cross validation. Then feature selection is performed using the Firefly Algorithm. Each ReLink dataset will be processed by each tree-based classification algorithm, namely Decision Tree, Random Forest and Deep Forest according to the results of the firefly feature selection. Performance evaluation uses the AUC value (Area under the ROC Curve). Research was conducted using google collab and the average AUC value generated by Firefly-Decision Tree is 0.66, the average AUC value generated by Firefly-Random Forest is 0.77, and the average AUC value generated by Firefly-Deep Forest is 0, 76. The results of this study indicate that the approach using the Firefly algorithm with Random Forest classification can work better in predicting software damage compared to other tree-based algorithms. In previous studies, tree-based classification with hyperparameter tuning on software defect prediction datasets obtained quite good results. In another study, the classification performance of SVM, Naïve Bayes and K-nearest neighbor with firefly feature selection resulted in improved performance. Therefore, this research was conducted to determine the performance of a tree-based algorithm using the firefly selection feature.

Pendekatan Algoritma Klasifikasi Machine Learning untuk Deteksi Penyakit Demensia

Early detection of dementia through the use of machine learning classification algorithms is important for providing appropriate interventions to patients. In this context, this study aims to compare the performance of several machine learning classification algorithms in detecting dementia using the attribute selection method. In the early stages, patient data including medical history, cognitive test results, and other attributes were collected as input, an attribute selection algorithm was used to select the most informative attribute subset in detecting dementia. The subset of attributes used as input for training machine learning classification models, several classification algorithms such as Extra Trees (ET), Linear Discriminant Analysis (LDA), Random Forest (RF) and Ridge. In this study, accuracy is used as the main metric to compare algorithm performance. The evaluation results show that the Random Forest (RF) algorithm produces the best performance with an accuracy of 91.56%. The Extra Trees (ET) algorithm has an almost comparable accuracy of 91.44%, while Ridge and Linear Discriminant Analysis (LDA) have an accuracy of 90.44% respectively. In the context of dementia detection, the performance of the Random Forest algorithm with the attribute selection method proved to be the best with an accuracy of 91.56%. These results indicate that the developed model is capable of recognizing complex patterns and relationships between features that are relevant to dementia status. The use of the attribute selection method also contributes to increasing the accuracy and efficiency of the classification algorithm.

Comparison of Performance of Classification Algorithms Using Standard Deviation-based Feature Selection in Cyber Attack Datasets

Supervised machine learning techniques are commonly used in many areas like finance, education, healthcare, engineering, etc. because of their ability to learn from past data. However, such techniques can be very slow if the dataset is high-dimensional, and also irrelevant features may reduce classification success. Therefore, feature selection or feature reduction techniques are commonly used to overcome the mentioned issues. On the other hand, information security for both people and networks is crucial, and it must be secured without wasting the time. Hence, feature selection approaches that can make the algorithms faster without reducing the classification success are needed. In this study, we compare both the classification success and run-time performance of state-of-the-art classification algorithms using standard deviation-based feature selection in the aspect of security datasets. For this purpose, we applied standard deviation-based feature selection to KDD Cup 99 and Phishing Legitimate datasets for selecting the most relevant features, and then we run the selected classification algorithms on the datasets to compare the results. According to the obtained results, while the classification success of all algorithms is satisfying Decision Tree (DT) was the best one among others. On the other hand, while Decision Tree, k Nearest Neighbors, and Naïve Bayes (BN) were sufficiently fast, Support Vector Machine (SVM) and Artificial Neural Networks (ANN or NN) were too slow.

#3836 DETERMINATION OF PREDICTORS OF OUTCOMES FOLLOWING VASCULAR ACCESS CONVERSION: WHICH PATIENTS BENEFIT MOST?

Abstract Background and Aims Conversion from central venous catheter (CVC) to arteriovenous fistula or graft (non-CVC) associates with favorable outcomes in hemodialysis (HD) patients. However, the “fistula first” paradigm has been debated and counter-proposed as “right access for the right patient at the right time and for the right reason”, indicating a shift towards precision medicine in vascular access care. It is unclearly how to classify patients into those suitable for conversion from CVC to non-CVC and which subgroups benefit most. Many factors (such as demographic, clinical and laboratory features) have been proposed to predict non-CVC failure after conversion and adverse outcomes. Due to confounding and potential biases it is difficult to shed light on which patients would benefit most from conversion. We investigated which factors are likely to be most helpful in the prediction of consequent benefit from conversion. Method We studied 54 potential predictors of incident HD patients undergoing conversion from CVC to non-CVC, between Jan. 2016 and Dec. 2019. Predictors included demographic and clinical variables such as comorbidities, drug history, and lab parameters. First, feature importance was assessed to separate weak from strong predictors using the Boruta algorithm. Second, important and tentative features were utilized in a subsequent machine learning workflow. Our main outcome was the predictive performance of different machine learning classification algorithms to predict re-conversion to central venous catheter and mortality within 1 year after conversion. Performance was quantified as accuracy, sensitivity, and Area under the receiver operating characteristics curve (AUC ROC). We compared insights from Machine learning algorithms to multivariate logistic regression models with selected inputs based on published literature (sex, age, BMI, diabetes, inflammation, blood pressure). Results After exclusion of patients with missing data, 38,151 out of 73,031 incident US HD patients were studied. Of these, 25,470 (66.8%) experienced no major adverse outcome within 1 year after access conversion, 3,683 (9.7%) underwent re-conversion, 7,779 (20.4%) did not survive the observational period, and 1,219 (3.2%) required re-conversion and did not survive the follow-up period. Post HD systolic blood pressure, history of a previous non-CVC that failed, and anthropometric characteristics (height, weight) had most weight in the prediction of re-conversion to CVC, according to the Boruta algorithm. Ethnicity and age were found to be the most important predictors of mortality. Classification based on the absence/presence of any major adverse outcomes resulted in a predictive accuracy between 0.70 and 0.54, depending on the respective algorithm. Sensitivity and ROCAUC had maxima of 0.58 and 0.69, respectively. However information gain by including all additional values had no remarkable effect on the predictive qualities (Table 1). Logistic regression for re-conversion in survivors only resulted in an accuracy of 0.78 and sensitivity of 0.29. Limiting predictors to 6 published predictors of our studied outcomes in the context resulted in an accuracy of 0.77 and sensitivity of 0.22 (Figure 1a and b). Conclusion Prediction of re-conversion and mortality within 1 year after catheter-to-arteriovenous access conversion is accurately feasible based on demographic and clinical features but discrimination of those benefiting most comes with low predictive accuracy. It remains reasonable to assume that not all patients will truly benefit from conversion, the inability to identify those that do not based on retrospective medical records data, again emphasizes “Fistula first” (if surgically feasible) and the need for investigation into molecular and genetic risk determinants.

Comparing Machine Learning and PLSDA Algorithms for Durian Pulp Classification Using Inline NIR Spectra.

The aim of this study was to evaluate and compare the performance of multivariate classification algorithms, specifically Partial Least Squares Discriminant Analysis (PLS-DA) and machine learning algorithms, in the classification of Monthong durian pulp based on its dry matter content (DMC) and soluble solid content (SSC), using the inline acquisition of near-infrared (NIR) spectra. A total of 415 durian pulp samples were collected and analyzed. Raw spectra were preprocessed using five different combinations of spectral preprocessing techniques: Moving Average with Standard Normal Variate (MA+SNV), Savitzky-Golay Smoothing with Standard Normal Variate (SG+SNV), Mean Normalization (SG+MN), Baseline Correction (SG+BC), and Multiplicative Scatter Correction (SG+MSC). The results revealed that the SG+SNV preprocessing technique produced the best performance with both the PLS-DA and machine learning algorithms. The optimized wide neural network algorithm of machine learning achieved the highest overall classification accuracy of 85.3%, outperforming the PLS-DA model, with overall classification accuracy of 81.4%. Additionally, evaluation metrics such as recall, precision, specificity, F1-score, AUC ROC, and kappa were calculated and compared between the two models. The findings of this study demonstrate the potential of machine learning algorithms to provide similar or better performance compared to PLS-DA in classifying Monthong durian pulp based on DMC and SSC using NIR spectroscopy, and they can be applied in the quality control and management of durian pulp production and storage.

Multiclass Classification of Paddy Leaf Diseases Using Random Forest Classifier

With increase in population, improving the quality and quantity of food is essential. Paddy is a vital food crop serving numerous people in various continents of the world. The yield of paddy is affected by numerous factors. Early diagnosis of disease is needed to prevent the plants from successive stage of disease. Manual diagnosis by naked eye is the traditional method widely adopted by farmers to identify leaf diseases. However, when the task involves manual disease diagnosis, problems like the hiring of domain experts, time consumption, and inaccurate results will arise. Inconsistent results may lead to improper treatment of plants. To overcome this problem, automatic disease diagnosis is proposed by researchers. This will help the farmers to accurately diagnose the disease swiftly without the need for expert. This manuscript develops model to classify four types of paddy leaf diseases bacterial blight, blast, tungro and brown spot. To begin with, the image is preprocessed by resizing and conversion to RGB Red, Green and Blue (RGB) and Hue, Saturation and Value (HSV) color space. Segmentation is done. Global features namely: hu moments, Haralick and color histogram are extracted and concatenated. Data is split in to training part and testing part in 70:30 ratios. Images are trained using multiple classifiers like Logistic Regression, Random Forest Classifier, Decision Tree Classifier, K-Nearest Neighbor (KNN) Classifier, Linear Discriminant Analysis (LDA),Support Vector Machine (SVM) and Gaussian Naive Bayes. This study reports Random Forest classifier as the best classifier. The Accuracy of the proposed model gained 92.84% after validation and 97.62% after testing using paddy disordered samples. 10 fold cross validation is performed. Performance of classification algorithms is measured using confusion matrix with precision, recall, F1- score and support as parameters.

Transforming Data with Ontology and Word Embedding for an Efficient Classification Framework

Transforming data into appropriate formats is crucial because it can speed up the training process and enhance the performance of classification algorithms. It is, however, challenging due to the complicated process, resource-intensive and preserved meaning of the data. This study proposes new approaches to building knowledge representation models using word-embedding and ontology techniques, which can transform text data into digital data and still keep semantic/context information of themselves in order to enhance modeling data later. To evaluate the effectiveness of the built models, a classification framework is proposed and performed on a public real dataset. Experimental results show that the constructed knowledge representation models contribute significantly to the performance of classification methods.

Addressing the class-imbalance and class-overlap problems by a metaheuristic-based under-sampling approach

The problem of imbalanced class distribution in real-world datasets severely impairs the performance of classification algorithms. The learning task becomes more complicated and challenging when there is also the class-overlap problem in imbalanced data. This research tackles these problems by presenting an under-sampling approach based on a metaheuristic method in which the under-sampling problem is mapped into an optimization problem. The proposed approach aims to select an optimal subset of the majority samples to handle the imbalanced and the class-overlap problems simultaneously while avoiding the excessive elimination of majority samples, especially in overlapped regions. The quality of the generated solutions is evaluated by a classifier and optimized in an evolutionary process. Unlike most existing under-sampling methods, the majority samples are not removed only from the overlapped regions; the classifier performance determines the desired regions for eliminating the majority samples. Extensive experiments conducted on 66 synthetic and 24 real-world datasets with different imbalance ratios and overlapping degrees and two large high-dimensional datasets show a significant performance improvement from the proposed method compared to the competitors.