feature selection

Stock price crash risk is a phenomenon in which stock prices are subject to severe negative and sudden adjustments. So far, different approaches have been proposed to model and predict  the  stock price crash risk, which in most cases have been the main emphasis on the factors affecting it, and often traditional methods have been used for prediction. On the other hand, using  Meta Heuristic Algorithms, has led to a lot of research in the field of finance and accounting. Accordingly, the purpose of this research is to model the Stock price crash risk of listed companies in Tehran Stock Exchange using firefly algorithm and compare the results with multivariate regression as a traditional method. Of the companies listed on the stock exchange, 101 companies have been selected as samples. Initially, 19 independent variables were introduced into the model as input property of the particle accumulation algorithm, which was considered as a feature selection method. Finally, in each of the different criteria for calculating the risk Stock price crash risk, some optimal variables were selected, then using firefly algorithm and multivariate regression, the stock price crash risk was  predicted  and results were compared. To quantify the Stock price crash risk, three criteria for negative skewness, high fluctuations and maximum sigma have been used. Two methods of MSE and MAE have been used to compare the methods. The results show that the ability of meta-meta-heuristic methods to predict the risk Stock price crash risk is not  generally higher than the traditional method of multivariate regression, And the research hypothesis was not approved.

In this study, an evaluation model is developed to assess the credibility of the loan applicants. The proposed model is a multicriteria decision making (MCDM) problem consisting of numerous criteria by integrating analytic hierarchy process (AHP) and genetic algorithm (GA). In the case of apparent consensus for several measures, the research clearly indicates that both quantitative and qualitative information must be employed in evaluating loan applicants. The AHP approach is widely used for MCDM in various scopes. In 2008 Lin et al proposed the adaptive AHP approach (A 3 )in order to decrease the number of steps for checking the inconsistency in the AHP model. The study presents a MCDM model by developing the new adaptive AHP approach (N_A 3 ) already proposed by Herrera-Viedma in 2004. The proposed model has led to fewer calculations, and less complexity. The model was applied to 200 clients in order to show its efficiency and applicability. A brief look at the implementation of the model showed that it is significantly valid in selecting clients with respect to the known criteria, besides decision making regarding the determination of the assessment factors.

This paper presents an approach that uses both genetic algorithm (GA) and fuzzy inference system (FIS), for feature selection for descriptor in a quantitative structure activity relationships (QSAR) classification and prediction problem. Unlike the traditional techniques that employed GA, the FIS is used to evaluate an individual population in the GA process. So, the fitness function is introduced and defined by the error rate of the GA and FIS combination. The proposed approach has been implemented and tested using a data set with experimental value anti-human immunodeficiency virus (HIV) molecules. The statistical parameters q2 (leave many out) is equal 0.59 and r (coefficient of correlation) is equal 0.98. These results reveal the capacity for achieving subset of descriptors, with high predictive capacity as well as the effectiveness and robustness of the proposed approach.

Dimensionality reduction is among the data mining process that is used to reduce the noise and complexity of features in various datasets. Feature selection (FS) is one of the most commonly used dimensionalities that reduces the unwanted features from the datasets. FS can be either wrapper or filter. Wrappers select subsets of the feature with better classification performance but are computationally expensive. On the other hand, filters are computationally fast but lack feature interaction among selected subsets of features which in turn affect the classification performance of the chosen subsets of features. This study proposes two concepts of information theory mutual information (MI). As well as entropy (E). Both were used together with binary cuckoo optimization algorithm BCOA (BCOA-MI and BCOA-EI). The target is to improve classification performance (reduce the error rate and computational complexity) on eight datasets with varying degrees of complexity. A support vector machine classifier was used to measure and computes the error rates of each of the datasets for both BCOA-MI and BCOA-E. The analysis of the results showed that BCOA-E selects a fewer number of features and performed better in terms of error rate. In contrast, BCOA-MI is computationally faster but chooses a larger number of features. Comparison with other methods found in the literature shows that the proposed BCOA-MI and BCOA-E performed better in terms of accuracy, the number of selected features, and execution time in most of the datasets.

Cloud Computing has drastically simplified the management of IT resources by introducing the concept of resource pooling. It has led to a tremendous improvement in infrastructure planning. The major goals of cloud computing include maximization of computing resources with minimization of cost. But the truth is that everything has a price and cloud computing is no different. With Cloud computing there comes a number of security concerns which need to be addressed. Cloud forensics plays a vital role to address the security issues related to cloud computing by identifying, collecting and studying digital evidence in cloud environment. The aim of the research paper is to explore the concept of cloud forensic by applying optimization for feature selection before classification of data on cloud side. The data is classified as malicious and non-malicious using convolutional neural network. The proposed system makes a comparison of models with and without feature selection algorithms before applying the data to CNN. A comparison of different metaheuristics algorithms- Particle Swarm Optimization, Shuffled Frog Leap Optimization and Fire fly algorithm for feature optimization is done based on convergence rate and efficiency.

The surge in internet usage has sparked new demands. Historically, specialized web crawlers were devised to retrieve pages pertaining to specific subjects. However, contemporary needs such as event identification and extraction have gained significance. Conventional web crawlers prove inadequate for these tasks, necessitating exploration of novel techniques for event identification, extraction, and utilization. This study presents an innovative approach for detecting and extracting events using the Whale Optimization Algorithm (WOA) for feature extraction and classification. By integrating this method with machine learning algorithms, the proposed technique exhibits improvements in experiments, including decreased execution time and enhancements in metrics such as Root Mean Square Error (RMSE) and accuracy score. Comparative analysis reveals that the proposed method outperformed alternative models. Nevertheless, when tested across various data models and datasets, the WOA model consistently demonstrated superior performance, albeit exhibiting reduced evaluation metrics for Wikipedia text data.

This paper focuses on a new hybrid machine learning model for classifying eye states from EEG signals by integrating traditional techniques with deep learning methods. Our Hybrid LSTM-KNN architecture employs KNN for classification and uses LSTM networks to extract features temporally. In addition, we perform extensive feature engineering, including statistical Z-test and IQR filtering, dimensionality reduction using PCA, and multivariate analysis to further model the performance. Moreover, an SVM-based unsupervised clustering approach is proposed to partition the EEG feature space, followed by ensemble learning in each cluster to improve accuracy and robustness. Using the EEG Eye State Dataset for the first assessment, the Hybrid LSTM-KNN model recorded an accuracy of 87.2% without PCA. Further improvements through statistical filtering outperformed initial expectations, achieving a 6% rise in performance to 89.1% after outlier removal, 89.1% with Z-test (σ = 3), and 88.3% with IQR (1.5x). After applying PCA along with ensemble learning post clustering, the final model exceeded expectations with an accuracy and F1 score of 96.8%, surpassing Ensemble Cluster-KNN and traditional models based on Ensemble Cluster-KNN, Logistic Regression, SVM, and Random Forest. The outcome demonstrates the robustness and noise-resilience of the model’s performance in practical real-time brain-computer interface and cognitive monitoring systems.