A Fast, Dynamic method to identify attributes sets using Corelation-Guided Cluster analysis and Genetic algorithm Techniques

Abstract

Dimensionality reduction is essential in both data mining and machine learning techniques. Reduction of the number of functions can be defined as "dimensionality." Computational expense and large amounts of data are making machine learning models challenging to use in the present environment. As the number of characteristics grows, the model becomes more dynamic, and overfitting becomes more likely. As training a machine learning model on a large number of features produces overreliance on the training data, model performance is low on real-world data, resulting in poor outcomes. Based on correlation-based cluster analysis and Genetic algorithm, this research offers a novel FS hybrid three-stage method. Two methodologies are developed in order to better pinpoint the target region of the third phase. These methodologies are a function clustering-based technique with minimal computational cost and a filter FS approach. The third stage is to find a subset of functions that provides an ideal mix of readability and functionality by way of a global readability evolutionary algorithm. Symmetrical insecurity-based removal, fast link-based clustering, and genetic algorithm: These stages are meant to increase the efficacy of a symmetrical insecurity-based removal procedure, a quick link-based clustering technique, and a genetic algorithm. For the sake of comparison, the suggested algorithm is contrasted to other FS algorithms found on openly accessible datasets in the physical world. The experimental findings shown that, with the least price of computation, the method can discover a successful function subset.