← Back to Publications List

A Hybrid Framework for Fair and Real-Time Crime Hotspot Prediction

Students & Supervisors

Student Authors
Abdullah Al Amin
Master of Science in Computer Science, FST
Supervisors
Firoz Ahmed
Professor, Faculty, FST

Abstract

Nowadays, crime and crime hotspots are beginning to be a thing in various cities in the world, because of which research on crime hotspot detecting and predicting is increasing. Machine Learning and Deep Learning techniques are gaining considerable attention in this field. Researches are mostly area-focused, which is a barrier for general data. The objective of this work is to find an effective real-time crime hotspot prediction by using unstructured data from major Bangladeshi online news articles, using hybrid features to improve crime prediction accuracy and apply fairness check in crime data. Which is done by scraping data from online newspapers, then processing it and training, evaluating a hybrid framework that’s built using NLP and a Random Forest Classifier to predict hotspots and check data fairness. This process helps to evaluate real time crime data, predict crime hotspot, provide temporal crime report also provide fairness audit reports. Evaluated on a custom dataset which contains 152 articles after sorting from 12 and 13 December 2025, the model achieved 82% accuracy with an AUC of 0.93. This addresses the real-world prediction and data fairness problems and provides a new hybrid model.

Keywords

Crime Crime Prediction Crime Hotspot Real-Time Prediction Hybrid Framework

Publication Details

  • Type of Publication:
  • Conference Name: International Conference on Electrical, Computer and Communication Technologies (ECCT 2026)
  • Date of Conference: 05/07/2026 - 05/07/2026
  • Venue: Dhaka International University, Dhaka, Bangladesh
  • Organizer: Dhaka International University
View on Publisher