Unsupervised Domain Adaptation for Crime Prediction Across Cities

Abstract

One of the most significant and pervasive issues in our society is a crime, and preventing it is a crucial task. An increasing crime factor leads to an imbalance in the constituency of a country. Crime prediction and forecasting is a challenging task for law enforcement agencies to prevent crimes in the future. In recent years, machine learning algorithms have been used to analyze crime data and provide useful insights to predict and prevent future crimes. In this paper, we propose a crime risk prediction and forecasting system using a sequential minimal optimization algorithm, a popular support vector machine algorithm that can be used for classification and regression tasks. We demonstrate the effectiveness of the SMO algorithm and LSTM model on a real-world crime dataset and compare its performance with other commonly used machine learning algorithms. Our results show that the SMO algorithm and LSTM model gives faster and more variety of visualizations for crime trend prediction and forecasting.