Enhancing monitoring of suspicious activities with AI-based and big data fusion

Abstract

This study provides an AI-based detection tool for the surveillance of suspicious activities using data fusion. The system leverages time, location, and specific data pertaining to individuals, objects, and vehicles associated with the agency. The study’s training data was obtained from Thailand’s military institution. The study focuses on comparing the efficiency between MySQL and Apache Hive for big data processing. According to the findings, MySQL is better suited for quick data retrieval and low storage capacity, while Hive demonstrates higher scalabilities for larger datasets. Furthermore, the study explores the practical utilization of web applications interfaces, enabling real-time display, analysis, and identification suspicious activity results. The web application, built with NuxtJS and MySQL, includes statistics charts and maps that show the status of suspicious items, cars, and people, as well as data filtering options. The system utilizes machine-learning algorithms to train the suspicious identification model, with the best-performing algorithms being the decision tree, reaching 98.867% classification accuracy.