Crime of theft prediction using Machine Learning K-Nearest Neighbour Algorithm at Polresta Bandar Lampung

Abstract

The era of the industrial revolution 4.0 is a time where cyber and physical technology collaborate. This study aims to predict the types of theft crimes that occur in the Bandar Lampung Police area with the K-Nearest Neighbor algorithm, evaluate the prediction results and profiling the prediction results carried out by Bandar Lampung Police investigators in efforts to prevent and handle criminal acts of theft in the jurisdiction of the Bandar Lampung Police Lampung. The approach was carried out using the quantitative method of the K-Nearest Neighbor algorithm using the Rapidminer application by utilizing 1671 police report data from the Bandar Lampung Police and a questionnaire survey method conducted on 49 police investigators from the Bandar Lampung Police. Data collection techniques are carried out in a valid and reliable manner as a support for predictive validity. Based on the results of the classification and questionnaire, it was found that the majority of victims of the crime of theft were adult men who did not have a job and lived in urban areas. It was found that the majority of thefts occurred in parking lots in urban areas on Monday morning where the perpetrators used tools and targeted moving objects by tampering with locks which caused losses of around 10-50 million rupiah. This type of theft is theft by weighting (CURAT) which applies to Article 363 of the Criminal Code. The prediction results show that the neighboring value (K) and the distribution ratio of training and testing data are K=3 and 7:3, respectively. Predictions using K values and data sharing ratios show a high level of accuracy, namely 99. 20%. The results of the questionnaire show results that are in line with the results of the classification with an accuracy rate of the actual data of 75. 7122%. So by increasing the understanding skills of Bandar Lampung Police investigators using technology to predict the crime of theft, the number of theft crimes can be reduced.

Implementation of Data Mining to Determine Public Interest in Automatic Motorcycles

“…K here is a parameter that determines the number of nearest neighbors to be considered (Adjani, Fauzia, & Juliane, 2023) (Arifuddin, Pinastawa, Anugraha, & Pradana, 2023). KNN is very effective for datasets that have a small number of dimensions (features), but can become less efficient as the number of dimensions increases due to the complexity of calculating the distance between data points (Hermawan & Prianggono, 2023) (Triani, Dar, & Yanris, 2023). KNN does not require a learning model in the initial phase, but requires more time in the testing phase because it needs to calculate the distance from new data points to all other data points that already exist in the dataset (A. W. Sari, Hermanto, & Defriani, 2023).…”