Design A Novel Feature Selection Technique With Machine Learning For Big Data Classification

Abstract

By combining feature selection techniques with machine learning classifiers, the effect of different methods of random sampling on the model's performance is looked at. This research presents a new feature selection method to fix issues caused by too many dimensions in data classification. The fields of machine learning and recognition of patterns have been looked into in great detail. Three primary categories of feature selection methods have been developed: filter, wrapper, and hybrid approaches. The effectiveness of technique through extensive experiments on diverse datasets, showcasing significant improvements in classification performance, computational efficiency, and model interpretability. The integration of feature selection strategies with machine learning classifiers allows an examination of the effects of different random sampling techniques on the model's performance. A unique feature selection technique for big dimensionality issues in data classification is presented in this study. The fields of pattern recognition and machine learning have been thoroughly investigated. There are three main types of feature selection techniques that have developed: hybrid, filter, and wrapper methods. They evaluate and compare every model's accuracy and performance, including the suggested model, Random Forest (RF), Logistic Regression, Decision Tree, and others. The optimal classifier is the model with the maximum accuracy. Achieving an accuracy of 0.926625, an F1-score of 0.955441, a precision of 0.981071, and a recall of 0.931116, the suggested model combined MLP and LSTM.