AI-powered Forensic Science - EvidenceOptimizer: Leveraging Hybrid Ensembled Models

Forensic science plays a pivotal role in criminal investigations by providing scientific methodologies to analyze and validate evidence. The EvidenceOptimizer framework enhances forensic science using a hybrid ensemble of deep learning and machine learning models. This study evaluates forensic evidence classification across multiple domains, including fingerprint analysis, ballistic matching, forensic text mining, and digital forensics. The ResNet-Inception Ensemble achieved 96.5% accuracy in fingerprint recognition, while the Siamese Network excelled in signature authentication with $\mathbf{9 7. 1} \boldsymbol{\%}$ accuracy. For forensic text classification, the BiLSTM-CNN Hybrid model reached an accuracy of 92.3%, outperforming traditional NLP-based methods. Ballistic evidence matching demonstrated high reliability, with SVM + Bayesian Network achieving 94.2% accuracy. The study also explores multimodal forensic analysis, where the Hybrid Ensemble (SVM + XGBoost +CNN) attained 95.1% accuracy in crime reconstruction. By integrating AI-driven forensic methodologies, EvidenceOptimizer enhances investigative accuracy, forensic efficiency, and legal transparency. The findings suggest that AI-powered forensic systems can significantly reduce false positives and negatives, improving forensic decision-making. Future research will explore real-time forensic evidence processing and federated learning for secure forensic AI applications. This study provides a transformative AI-based approach to forensic science, ensuring greater accuracy and reliability in criminal investigations.