<![CDATA[Crime is a phenomenon that needs to be understood and predicted to reduce victimizations and improve the efficiency of investments in personnel and equipment. Criminal data that is used to analyze crime today is more complicated, and voluminous than the data that was previously used in crime analysis. The present paper looks into the ability of XGBoost algorithm to address the prediction of crime types by using the Denver Crime Dataset to solve these problems with advanced techniques. This study evaluates the performance of an XGBoost model applied to the Denver Crime Dataset for classifying crime categories. Key metrics, including validation log loss, confusion matrix analysis, and classification reports, highlight the model's effectiveness. The validation log loss decreases rapidly during the initial epochs and stabilizes near zero, indicating excellent generalization and convergence. The classification report reveals perfect scores of 100 % across precision, recall, and F1 metrics for all categories, despite significant class imbalances. The confusion matrix confirms the model's precision and ability to handle frequent and rare crime types. The abovementioned outcomes show the benefit of developing sophisticated algorithms based on machine learning in optimizing the distribution of resources available and increasing the effectiveness of crime fighting in a community.]]>
Copyrights © 2025
