NOVEL METHODOLOGIES FOR PREDICTIVE ANALYSIS IN CRIME DATA OVER ONLINE SOCIAL MEDIA

Abstract

Social media platforms have become integral for communication and information exchange, yet they also pose challenges such as cybercrime and hacking. The escalating incidents of crimes on platforms like Twitter necessitate proactive measures, including crime prediction. This research employs a comprehensive approach, utilizing Twitter data for crime prediction through data preprocessing and feature extraction techniques. Techniques such as Bag of Words, Glove, TF-IDF (Term Frequency-Improved Document Frequency), and feature hashing are employed, with feature selection using a Modified Tree Growth Algorithm (MTGA) and clustering via Fuzzy Manta Ray Foraging (FMRF). The crime detection is performed using a hybrid Wavelet Convolutional Neural Network with World Cup Optimization (WCNN-WCO). The proposed method outperforms existing ones in terms of precision, accuracy, F1 measure, and recall, addressing the rising social issue of social media crimes. Furthermore, the study introduces DAC-BiNet, a robust Deep Attention Convolutional Bi-directional Aquila Optimal Network, specifically tailored for crime detection on the Twitter platform. The model undergoes a multi-stage process involving pre-processing, feature extraction, and clustering through Possibilistic Fuzzy LDA (Latent Dirichlet Allocation). Experimental results demonstrate the effectiveness of DAC-BiNet (Deep Attention Convolutional Bi-directional Aquila Optimal Network), achieving increased accuracy, precision, recall, specificity, and F1 score. Additionally, the paper explores the application of Apache Pig with Hadoop in large scale crime data analysis. Utilizing incident-level crime data, the study showcases the efficacy of Apache Pig in analyzing vast datasets, aiding decision-makers, policymakers, and governments in minimizing crime. In conclusion, these research works collectively highlight the significance of technology-driven approaches in addressing and mitigating the complex issues surrounding crime, be it on social media platforms or in large-scale datasets. The findings provide valuable insights for law enforcement, policymakers, and governments to make informed decisions and formulate effective strategies in the realm of crime prevention and control.