BIASED CLUSTERING AND FEATURE SELECTION FOR ENHANCED PREDICTIVE MODELING IN SOCIAL MEDIA INFLUENCE MAXIMIZATION

Abstract

Social Homophily and Influence Predictive modeling for Social Recommendation is a vital tool in various fields, aiding in decision-making and forecasting future trends. This research delves into an innovative approach that combines biased clustering and feature selection techniques to enhance the accuracy and relevance of predictive models. The methodology unfolds in two pivotal steps. In this research we use three types of datasets are Facebook, Instagram, Youtube for Predictive modeling.After dataset collection data preprocessing is executed using a novel adaptation of the Biased Renovate K-Means Clustering. After data preprocessing we use Biased Bat Algorithmwith an Improved Extra Tree Classifier for feature selection. This approach integrates the ability of met heuristic optimization (Biased Bat Algorithm) with a robust feature evaluation technique (Improved Extra Tree Classifier). The incorporation of bias in feature selection allows for the prioritization of features based on domain knowledge or research objectives, enriching the modeling process with contextual relevance. The synergy between biased clustering and feature selection augments the efficiency and effectiveness of predictive modeling. Facebook, Instagram, Youtube stands out with exceptional performance accuracy of 95.21%, 96.11%, 97.54%. By tailoring both data preprocessing and feature selection to specific criteria, the resulting models are more attuned to the underlying patterns in the data, thus enhancing prediction accuracy.

Keywords:Biased Clustering, Biased Bat Algorithm, Feature Selection, Improved Extra Tree Classifier Renovate K-Means Clustering