Optimizing Computational Efficiency in Feature Selection for Machine Learning Models: A Study Crime Detection Based on Criminal Data

In this study, we embarked on the journey to optimize the computational efficiency of feature selection techniques for machine learning models, with a focused case study on crime detection using criminal data. Through a comprehensive analysis of various aspects of the data, from preprocessing to feature selection and model evaluation, valuable insights were gained. The results of our investigation underscored the significance of feature selection in enhancing both the accuracy and computational time of machine learning models. By comparing three classification models—Random Forest, Naïve Bayes, and AdaBoost. We observed that the performance varied significantly across different models and feature subsets. This research aims to explore the impact of feature selection on the accuracy and computational efficiency of machine learning models applied to crime detection tasks. By leveraging a case study approach, we seek to uncover the interplay between various feature selection techniques and classification models. The research encompasses a comprehensive five-step methodology, including data collection, preprocessing, feature selection, model classification, and performance evaluation. We adopt mutual information-based feature selection to distill relevant attributes from the dataset. The results of our research show that the random forest model has the best accuracy value among other models when using feature selection which is 0.99. In terms of computation time, random forest has a time of 66 seconds.