Risk assessment for commercial crime and legal protection based on modularity-optimized LM and SCA-optimized CNN

Abstract

With the Internet of Things (IoT) making significant strides in recent years, the challenges associated with data collection and analysis have emerged as a pressing concern in public security. When employed to tackle extensive criminal networks, the conventional deep learning model encounters issues such as heightened computational complexity, sluggish operational efficiency, and even system failures. Consequently, this research article introduces an intricately devised framework for detecting commercial offenses, employing a modularity-optimized Louvain-Method (LM) algorithm. Additionally, a convolutional neural networks (CNN)-based model is formulated to determine the feasibility of extending legal aid, wherein feature transformation is facilitated by utilizing TFIDF and Word2vec algorithms aligned with diverse legal text corpora. Furthermore, the hyper-parameter optimization is accomplished using the sine cosine algorithm (SCA), ultimately enabling the classification of relevant legal guidance. The experimental outcomes comprehensively affirm the exceptional training effectiveness of this model. The commercial crime identification model, grounded in modular optimization as proposed in this article, adeptly discerns criminal syndicates within the commercial trading network, achieving an accuracy rate exceeding 90%. This empowers the identification of such syndicates and bestows the judicial sphere with pertinent legal insights.