Analisis Prediksi Tindak Pidana Pencurian Dengan Metode Klasifikasi Algoritma K-Nearest Neighbor di Polresta Bengkulu, Polda Bengkulu

This research is driven by the phenomenon of high crime rates, especially theft. A method is needed that can assist the police in planning and implementing a more optimal and targeted patrol strategy. Data mining can be applied to crime data, especially theft crime data, to obtain information that can be used as a basis for decision making in carrying out patrols by the Police. Machine learning algorithms can be used to classify types of theft crimes based on characteristics and predict the possibility of future theft crimes based on influencing factors. Data mining is a logical process to find information that is very useful as a supporting tool in decision making. K-Nearest Neighbor (KNN) is a classification method and uses the CRISP-DM (Cross Industry Standard Process for Data Mining) framework in pulling information from a collection of datasets. This study uses a population in the form of Police Reports (LP) of Theft Victims that occurred in the Bengkulu Police Resort which occurred from 2020 to 2025. The K-Nearest Neighbor (KNN) model shows a fairly high level of reliability in predicting the time of theft, the age of the victim, and the type of item stolen at Bengkulu Police Resort. The KNN model was able to predict the time of incidents, victim age, and type of stolen items with high accuracy — 87.35%, 82.41%, and 88.43% respectively. Based on the results of this study, the application of machine learning in predictive policing can be implemented more effectively than conventional patrol methods. Therefore, the KNN prediction model developed in this study is recommended to be applied in the police patrol system, especially at Bengkulu Police Resort, to improve the effectiveness of surveillance and crime prevention in the jurisdiction of Bengkulu City.