Penerapan Model Prediksi Untuk Diintegrasikan Dalam Program Analisis Kerja Personel Dalam Mendukung Peningkatan Perencanaan Strategis dan Operasional Kepolisian

di Polresta Bandar Lampung Polda Lampung

Abstract

In the Indonesian context, the application of predictive technology in the police sector still faces various challenges. Some of them are low accuracy, in-date, and difficulty accessing data that results in government bureaucratic inefficiency.  Bandar Lampung City is one of the regions in Indonesia with a high crime rate. Data from the Bandar Lampung Police shows that theft cases dominate crime reports in the area. In this study, the improvement of ML model performance is expected to provide more accurate predictions and support police officers in designing more appropriate strategies to tackle theft crimes This study uses two popular Machine Learning (ML) models, namely K-Nearest Neighbors (k-NN) and Naïve Bayes (NB), to analyze and predict theft crimes in the jurisdiction of the Bandar Lampung Police. The approach is carried out using a quantitative method of algorithm k-Nearest Neighbor and Naive Bayes using the Rapidminer application by utilizing 1671 data from the Bandar Lampung Police police report and a survey method with questionnaires. The data collection technique is carried out validly and reliably, then the police report data will be used for prediction and the questionnaire data will be used to support the validity of the prediction. Based on the results of comparative research conducted using the K-NN model and the Naive Bayes model, it is known that the k-NN model on theft victim data based on the type of theft that occurred is able to predict by 98.80% and for the Naive Bayes model is able to predict by 99.85%. And for suspect data in the k-NN model, it is predicted to be 70.00% while the Naive Bayes model predicts 88.00%. In predicting theft incidents in Bandar Lampung, the selection of the Naive Bayes (NB) model proved to be much more effective and had a very high accuracy compared to K-Nearest Neighbors (K-NN). Based on the test results, the Naive Bayes model provides a prediction accuracy of 99.85%, which is much better compared to K-NN which may not achieve the same level of accuracy.