A SYSTEMATIC REVIEW OF AI-DRIVEN MULTI-DISEASE EPIDEMIOLOGICAL MODELLING: METHODOLOGICAL GAPS AND RESEARCH DIRECTIONS

This study presents a systematic and engineering-oriented review of artificial intelligence (AI)-driven epidemiological forecasting models published between 2012 and 2022. A structured search was conducted across Scopus, Web of Science, PubMed, and IEEE Xplore using predefined Boolean search strings combining SEIR modelling, machine learning, deep learning, and multidisease forecasting terms. Following PRISMA 2020 guidelines, 236 records were identified, screened to 104 articles, and 18 studies met the final inclusion criteria based on methodological rigor, validation transparency, and relevance to predictive epidemic modelling. Quantitative synthesis indicates that 83% of reviewed studies focus on single-disease forecasting, only 11% incorporate co-infection dynamics, and fewer than 20% systematically integrate environmental or policy control variables. Comparative reporting across selected studies suggests that AI-based models reduce RMSE by approximately 12–18% relative to classical SEIR implementations, though heterogeneity precluded formal meta-analysis. Additionally, interpretability mechanisms such as SHAP remain underutilised in epidemiological contexts. The review identifies reproducible methodological gaps and proposes a mathematically structured hybrid SEIR-AI framework incorporating cross-pathogen interaction matrices, environmental drivers, and policy control inputs. The study contributes an engineering blueprint for scalable, interpretable multi-disease forecasting architectures suited to resource-constrained environments. Limitations include restricted sample size (n=18) and absence of statistical meta-aggregation.