Hybrid Machine Learning and Nature-Inspired Optimization for Robust and Accurate Product Recommendations

Abstract The rapid evolution of e-commerce and consumer behaviour demands intelligent and accurate product recommendation systems. This research presents a dual-path hybrid framework that integrates advanced machine learning and quantum-inspired techniques for high-precision, relevance-based product recommendation. Initially, raw datasets undergo pre-processing and noise removal to ensure data quality, followed by comprehensive product characteristic extraction. The first pipeline employs Fisher's Linear Discriminant Analysis (LDA) for feature selection, and Hierarchical Agglomerative Clustering (HAC) combined with attribute-based clustering to group similar products. Prediction is performed using a Stacked DenseNet121 model, and relevant product recommendations are generated through an attention-based Multilayer Perceptron (MLP). Outcome analysis is conducted to evaluate prediction performance and validate product relevance. The second pipeline emphasizes feature extraction using Multivariate Additive Independent Component Analysis (MAICA), followed by hybrid clustering with HAC and K-Means algorithms. Extracted features are optimized through a Multi-Heuristic BAT Optimization Algorithm to enhance classification accuracy. The refined data is then processed using a Deep Multilayer U-Net Classifier to perform product prediction. Finally, a ranking-based recommendation approach prioritizes products based on relevance and predicted demand. The integration of traditional deep learning models with hybrid clustering and nature-inspired optimization techniques ensures superior accuracy, robustness, and scalability. Experimental results reveal that the proposed dual-stream approach significantly improves Mean Squared Error (MSE), R² score, and overall prediction accuracy compared to conventional models. This intelligent recommendation framework can be effectively utilized in real-world e-commerce environments to offer users more accurate, relevant, and timely product suggestions.