Model to Predict the Probability of a Criminal Event in Peru Using Machine Learning Algorithms

Insecurity in the Callao (province of Peru) region has intensified in recent years, despite the presence of strategically distributed police stations. This issue is influenced by a lack of integration of advanced technologies in crime prevention, limiting the effectiveness of security strategies. To address this situation, the present study proposes a predictive model based on Machine Learning algorithms applied to the crime context of Callao-Peru. The methodology includes three main stages: Analysis of relevant features, model design, and results validation. In the analysis, eight variables (F1 to F8) associated with criminal patterns were evaluated, while the validation phase compared the performance of the algorithms in terms of accuracy (ACC), precision, F1-score, recall, and AUC. The results show that the Decision Tree algorithm had the highest overall accuracy, achieving an ACC of 0.97, followed by Gradient Boosting (ACC = 0.93) and Random Forest (ACC = 0.92), while K-Nearest Neighbors had the lowest performance (ACC = 0.72). These findings position the Decision Tree as the most effective tool for predicting high-risk crime areas in Callao-Peru.