MACHINE LEARNING IN CRIME HOTSPOT FORECASTING: OPTIMIZING RESOURCE DISTRIBUTION FOR PUBLIC SAFETY

Predicting crime hotspots using Machine Learning has emerged as a promising

technique for boosting law enforcement strategies. This assignment examines a dataset of over

1.7 million historical crime records, utilizing temporal and spatial analysis techniques to

identify patterns in crime occurrences. The crime facts are labeled into three levels—low,

medium, and immoderate—based mostly on crime rates. Several tool-studying algorithms have

been accomplished and compared, which include Decision Trees, Random Forest, Naive Bayes,

K-Nearest Neighbor (KNN), Logistic Regression, Linear Discriminant Analysis (LDA),

Quadratic Discriminant Analysis (QDA), and Gradient Boosting Trees. The Overall

performance metrics of the model were assessed using metrics such as accuracy, confusion

matrices, and Unweighted Average Recall (UAR). The Random Forest classifier exhibited

superior average overall performance, achieving immoderate accuracy and UAR scores for

every balanced and imbalanced dataset. Additionally, a "God Test" was modified into brought

to evaluate the model`s predictive energy using historical facts, further validating its

effectiveness on unseen datasets. The findings highlight the functionality of tool studying to

assist facts-driven selections in law enforcement, allowing optimized police patrols and better

useful resource allocation to reduce crime correctly.