Domain Enhanced Pre-processing for Disease-Aware Recipe Recommendation Systems

Abstract

Data pre-processing is an important stage in machine learning, especially for domain-specific applications like personalised nutrition and disease-aware recommendation systems. This study clarifies a hybrid pre-processing framework for disease-specific recipe recommendation that combines general machine learning techniques (missing value handling, feature scaling, encoding, and outlier detection) with domain-specific enhancements (ingredient text normalisation, nutritional profiling, and disease-aware filtering). The suggested strategy promotes dietary compliance for illnesses like diabetes and ulcerative colitis by integrating ingredient appropriateness scores and health-related limitations. The experimental results show that these pre-processing procedures greatly increase recommendation accuracy and personalisation, lowering bias and improving the model's capacity to create health-conscious meal recommendations. This study offers important insights for health informatics, AI-driven personalised nutrition, and machine learning-based food recommendation systems, emphasising the importance of strong data pre-processing pipelines in specialised ML applications.