EDGE ASSISTED CRIME PREDICTION AND EVALUATION FRAMEWORK FOR MACHINE LEARNING ALGORITHMS

Abstract

In current days, With the rapid growth of global populations, particularly in major cities, new challenges
in public safety regulation and optimization have emerged. This paper introduces a strategy for predicting crime
occurrences in urban areas based on historical events and demographic data. The proposed framework leverages
machine learning algorithms deployed at the network edge to analyze four specific types of crimes: murder, rapid
trial, repression of women and children, and narcotics. Through a comprehensive study and implementation
process, we have developed a visual representation of crime distribution across various regions.The methodology
involves selecting, assessing, and implementing different Machine Learning (ML) models to predict criminal risk for
specific time intervals and locations. Techniques such as Decision Trees, Neural Networks, K-Nearest Neighbors,
and Impact Learning are employed, with performance comparisons based on data processing and modification
approaches. The Decision Tree algorithm achieved the highest accuracy at 81% in crime prediction. Our findings
indicate that machine learning techniques can effectively predict criminal events, thereby contributing to improved
public security.