Dementia is a neuropsychiatric brain disorder that usually occurs when one or more brain cells stop working partially or at all. Diagnosis of this disorder in the early phases of the disease is a vital task to rescue patients’ lives from bad consequences and provide them with better healthcare. Machine learning methods have been proven to be accurate in predicting dementia in the early phases of the disease. The prediction of dementia depends heavily on the type of collected data which usually are gathered from Normalized Whole Brain Volume (nWBV) and Atlas Scaling Factor (ASF) which are normally measured and corrected from Magnetic Resonance Imaging (MRIs). Other biological features such as age and gender can also help in the diagnosis of dementia. Although many studies use machine learning for predicting dementia, we could not reach a conclusion on the stability of these methods for which one is more accurate under different experimental conditions. Therefore, this paper investigates the conclusion stability regarding the performance of machine learning algorithms for dementia prediction. To accomplish this, a large number of experiments were run using 7 machine learning algorithms and two feature reduction algorithms namely, Information Gain (IG) and Principal Component Analysis (PCA). To examine the stability of these algorithms, thresholds of feature selection were changed for the IG from 20% to 100% and the PCA dimension from 2 to 8. This has resulted in 7×9 + 7×7= 112 experiments. In each experiment, various classification evaluation data were recorded. The obtained results show that among seven algorithms the support vector machine and Naïve Bayes are the most stable algorithms while changing the selection threshold. Also, it was found that using IG would seem more efficient than using PCA for predicting Dementia. These promising results open the door to a new era of early prognosis of Alzheimer’s Disease and Related Dementias (ADRD).

Human resources (HR) of various groups (e.g., ethnic minority or majority) and their quality play a crucial role in developing and promoting economic and social progress in every region. However, current methods of quality assessment, e.g., surveys, have not provided data-driven insight for policymakers to design targeted interventions. Machine learning is one of the emerging technologies that could analyze complex datasets to support data insight for policymakers in sustainable economic development. This study proposes a framework to predict the quality of HR from ethnic minority community by using various machine learning techniques (K-nearest neighbors, multilayer perceptron, gradient boosting, and voting classifier). To achieve the best model, two techniques for feature selection (recursive feature elimination and extra trees) are employed. In the experiments, the ethnic minority HR data has been used to conduct. Experimental results show that the gradient boosting consistently outperformed other models across feature selection techniques (≥0.99). The findings from this study enhance prediction methods for HR and provide valuable insights for policymakers to develop effective policies for ethnic minority communities.

The effectiveness and efficiency of a machine learning model can be improved by feature selection, especially for high-dimensional datasets such as in cybersecurity. The proposed approach utilizes an enhanced version of the Rainbow agent with a memory storage structure. The suggested approach is assessed using two benchmark datasets namely RT-IoT2022 which is targeted towards IoT network security and the Android Malware Detection dataset which is meant for mobile security. The specification of the reinforcement learning model has been trained for 20 epochs and it is progressively enhanced through feature subsets to enhance classification accuracy. The results show that the AUC scores continuously increase were the one for RT-IoT2022 achieves 0.91 and Android at 0.93. Three well-known classifiers XGBoost, Random Forest and multi-layer perceptron (MLP) are used to test the power of the selected features. The outcome evaluation on RT-IoT2022 dataset shows that Random Forest achieved maximum accuracy (99.48%), followed by XGBoost (99.16%), while MLP secured 94.04% accuracy. In the Android malware dataset, XGBoost model gave the best accuracy of 89.50%, followed closely by Random Forest with 87.00% and MLP with 86.50%. This clearly shows that reinforcement learning based feature selection enhances accuracy and reduces computation. The research emphasizes utilizing dynamic feature selection in any cyber security application. The future will experiment with incorporating deep reinforcement learning as well as hybrid selection.

Sentiment analysis of students’ feedback using machine learning algorithms has emerged as a valuable tool for understanding students’ sentiments and improving educational outcomes. Currently, existing systems use frequency-based methods for feature selection (e.g., Term Frequency-Inverse Document Frequency (TF-IDF) and Bag of Words (BoW)) not to capture the subtleties of emotions expressed in student feedback and do not provide insights into the specific concerns of students via topics or themes. In this study, we propose the Student Sentiment from Feedback (SSF) framework, which includes four main procedures: pre-processing, feature selection, classification, and theme finding. The SSF framework classifies student sentiments and subsequently groups feedback into themes using semantic networks based on word co-occurrence. Our innovative feature selection approach combines TF-IDF with sentiment-based features derived from SentiWordNet and intensifiers, creating a robust feature vector that enhances the dataset’s richness and improves classification accuracy and robustness. In the experiments, we utilize a public dataset from Kaggle, applying our proposed method and various machine learning models (e.g., k-nearest neighbor, decision tree, random forest, multilayer perceptron, support vector machine, gradient boosting, and extreme gradient boosting). The experimental results show that our concatenated features achieve the highest accuracy across all machine learning models (greater than 0.82). Our study demonstrates the efficacy of this hybrid feature selection method, contributing to better understanding and decision-making in educational settings.

The increasing amount of IoT devices increases the size of network traffic data, causing an increase in the incidence of security breaches in IoT networks. Cybercriminals have developed malware to compromise the security of sensitive data, among other cyber threats. In the presence of inadequate and robust security mechanisms, sensitive data is prone to vulnerability. Hence, protecting data in the IoT environment is becoming a mandatory task. Various approaches have addressed malware detection using network data features. However, there is still room for improvement in developing superior techniques and utilizing more comprehensive datasets. This paper presents a novel lightweight ensemble voting classifier to detect malware traffic by deploying the best possible network data. The merits of the correlation coefficient and Opposition-Based Crow Search Algorithm (OCS) have been leveraged to compute the best possible features. Another advantage of this proposed experiment is its focus on a dataset tailored to malware traffic features. This focus enables highly accurate malware detection. After feature selection using OCS, the proposed malware classifier is trained and validated with both 5-fold and 10-fold cross-validation techniques. The tested results confirm that the presented malware classifier performs best using a minimal feature set, which is highly advantageous for IoT networks due to resource constraints.

Information and Communication Technologies (ICT) affect all aspects of our daily lives. Using them is considered a symbol of modernization and social advancement. The global expansion and interconnection of ICT offers a significant opportunity to promote the advancement of humanity, bridge the digital gap and promote the growth of societies built on knowledge. In this study, we analyzed and identified the most influential factors in the adoption of ICT in education from the data set called “Final Survey-Digital Inclusion Teachers” of the Plurinational State of Bolivia, which consists of 871 instances and 189 columns. We performed feature selection by carefully combining the results of three feature selection methods: filter (chi-square, ANO-VA and mutual information), wrapper (RFE) and intrinsic (Classification And Regression Trees, Random Forest, Gradient Boosting and XGBoost). The results demonstrated that a teacher's motivation for curricular planning that includes ICT, teaching experience and the institutional environment are key factors in the adoption of these technologies in education. Furthermore, we identified that the Random Forest algorithm is the most appropriate for analyzing and predicting the adoption of ICT in education, we affirmed this after this algorithm obtained the highest values in four of the six metrics evaluated: a sensitivity of 77.7%, an F1 Score of 77.9%, a Cohen's Kappa coefficient of 60.8% and a Jaccard Score of 64.3%. These results suggest that Random Forest is the most effective algorithm to analyze the factors related to the adoption of ICT in educational environments.

Operational data is always huge. A preprocessing step is needed to prepare such data for the analytical process so the process will be fast. One way is by choosing the most effective features and removing the others. Feature selection algorithms (FSAs) can do that with a variety of accuracy depending on both the nature of the data and the algorithm itself. This inspires researchers to keep on developing new FSAs to give higher accuracies than the existing ones. Moreover, FSAs are essential for reducing the cost and effort of developing information system applications. Merging multiple methodologies may improve the dimensionality reduction rate retaining sensible accuracy. This research proposed a hybrid feature selection algorithm based on ScC and forward selection methods (ScCFS). ScC is based on stability and correlation while forward selection is based on Random Forest (RF) and Information Gain (IG). A lowered subset generated by ScC is fed to the forward selection method which uses the IG as a decision criterion for selecting the attribute to split the node of the RF to generate the optimal reduct. ScCFS was compared to other known FSAs in terms of accuracy, AUC, and F-score using several classification algorithms and several datasets. Results showed that the ScCFS excels other FSAs employed for all classifiers in terms of accuracy except FLM where it comes in second place. This proves that ScCFS is the pioneer in generating the reduced dataset with remaining high accuracies for the classifiers used.

Diabetes Mellitus (DM) is a frequent condition in which the body's sugar levels are abnormally high for an extended length of time. It is a major cause of death with high mortality rates and the second leading cause of total years lived with disability worldwide. Its seriousness comes from its long-term complications, including nephropathy, retinopathy, and neuropathy leading to kidney failure, poor vision and blindness, and peripheral sensory loss, respectively. Such conditions are life-threatening and affect patients’ quality of life. Therefore, this paper aims to identify the most relevant features in the diagnosis of DM and identify the best classifier that can efficiently diagnose DM based on a set of relevant features. To achieve this, four different feature selection methods have been utilized. Moreover, twelve different classifiers that belong to six learning strategies have been evaluated using two datasets and several evaluation metrics such as Accuracy, Precision, Recall, F1-measure, and ROC area. The obtained results revealed that the correlation attribute evaluation method would be the best choice to handle the task of feature selection and ranking for the considered datasets, especially when considering the Accuracy metric. Furthermore, MultiClassClassifier would be the best classifier to handle Diabetes datasets, especially when considering True Positive, precision, and Recall metrics.

In the realm of financial security, the detection and prevention of credit card fraud has become paramount. With the ever-increasing reliance on digital transactions, the risk of fraudulent activities targeting credit card systems has grown significantly. To combat this, sophisticated techniques are required to swiftly identify and mitigate potential threats. Machine learning, a cornerstone of modern data analysis, has emerged as a powerful tool in this pursuit. By leveraging vast datasets and employing advanced algorithms, machine learning enables the automated scrutiny of transactions, distinguishing between legitimate and fraudulent activities with remarkable precision. This paper introduces an intelligent method for credit card fraud detection that relies on Competitive Swarm Optimization (CSO) and Random Weight Network (RWN). Additionally, the system includes an automated hybrid feature selection capability to identify the most pertinent features during the detection process. The experimental outcomes validate that this system can attain outstanding results in G-Mean, RUC, and Recall values.

By everyday development of storage and communicational and electronic media, there are significant amount of information being collected and stored in different forms such as electronic documents and document databases makes it difficult to process them, properly. To extract knowledge from this large volume of documental data, we require the use of documents organizing and indexing methods. Among these methods, we can consider clustering and classification methods where the objective is to organize documents and to increase the speed of accessing to required information. In most of document clustering methods, the clustering is mostly executed based on word frequency and considering document as a bag of words. In this essay, in order to decrease the number of features and to choose basic document feature, we use background knowledge and word clustering methods. In fact by using WordNet ontology, background knowledge and clustering method, the similar words of documents are clustered and the clusters with the number of words more than threshold are chosen and then their frequency of words is accepted as the effective features of document. The results of this proposed method simulation shows that the documents dimensions are decreased effectively and consequently the performance of documents clustering is increased.