CRIME HOTSPOT PREDICTION

CRIME HOTSPOT PREDICTION

Sejal Dhondge 1, Fiza Shaikh 2, Alisha Shaikh 3, Prof. Bhagyashri Vyas4

1Department of Computer Engineering Dr. D. Y. Patil School of Engineering, Lohegoan Savitribai Phule Pune University, Pune Maharashtra, India

sejaldhondge123@gmail.com1

2Department of Computer Engineering Dr. D. Y. Patil School of Engineering, Lohegoan Savitribai Phule Pune University, Pune Maharashtra, India

fizashaikh23400@gmail.com2

3Department of Computer Engineering Dr. D. Y. Patil School of Engineering, Lohegoan Savitribai Phule Pune University, Pune Maharashtra, India

alishashaikh8786@gmail.com3

4Department of Computer Engineering Dr. D. Y. Patil School of Engineering, Lohegoan Savitribai Phule Pune University, Pune Maharashtra, India

bhagyashri.vyas@dypic.in4

Abstract:
Crime is a serious problem that affects many people in society. Crime prevention and reduction is a top priority for many countries. Given the limited resources for policing and reducing crime, it is important to identify effective strategies to optimize available resources. For this problem, crime gold estimation has been made before. Crime hotspot forecasting uses historical data to identify areas prone to future crime.However, the final method for crime hotspot prediction is to use only crime-related misinformation to identify crime hotspots and ignore the predictive power of other information such as public information or social media. In this article, we present Crime, a platform for predicting and visualizing crime hotspots based on the integration of different types of data. Our platforms regularly collect general and social data as well as crime data from across the web. It also provides useful features from data collected based on statistics and language analysis. Finally, it identifies crime points using the subtraction method and provides a view of hotspots on the interactive map.Crime forecasting is crucial to inform policing strategy and improve crime prevention and operations. Machine learning is now a predictive subject. But many studies have compared different types of machine learning to predict crime. This article uses real-time data on public crime from 2015 to 2018 in an area of a major coastal city in eastern China, based on observational data, to evaluate the predictive ability of various machine learning algorithms. Results from crime scene data alone show that the GEO OLD model outperforms CNNs, Arbitrary Wood, Support Vector Machines, Naive Bayes, and Convolutional Neural Networks.Additionally, erection topographic data related to point of interest (POIs) and public road network density are entered into the tGEO OLD model as covariates. Models producing terrain covariates have been found to have better predictive performance than traditional models based on false data on crime rates. Therefore, predicting future crime should use unbiased information about crime and crime-related covariates. Not all machine learning algorithms are equal in predicting crimes. Crime forecasting is crucial to inform policing strategy and improve crime prevention and operations.Machine learning is now a predictive subject.

Keywords: Face Detection, CNN, Geo-Fencing