Optimizing Crime Classification: HQFNN with Triangulation Topology Aggregation Optimizer

Analyzing crime activities in both spatial and temporal hotspots becomes a critical step to enhancing public safety and optimizing resource allocation in policing. This paper thus proposes a novel Triangulation Topology Hierarchical Quantum Fuzzy Network (TTHQFN) for crime type detection. Pre-processing includes outlier removal using the 3σ approach and also Min-Max standardization. Spatial and temporal features are extracted from Steerable Transformers, which captures complex dependencies with rotationally equivariant patterns well across spatial and temporal domains. Crime type is detected using a Hierarchical Quantum Fuzzy Neural Network (HQFNN) optimized with Triangulation Topology Aggregation Optimizer (TTAO) to obtain enhanced classification accuracy from fuzzy logic as well as the principles of quantum computing. Hyper parameters are further tuned and the errors are minimized to achieve optimal training of the network. The crime pattern and trend identification are achieved through the proposed framework, hence facilitating accurate crime-type prediction and classification. Results evaluation with the Chicago crime dataset showed high classification accuracy, precision, recall, f1-score of about 99%, and efficiency in computation time with 10s.