A data-driven intelligent decision support system that combines predictive and prescriptive analytics for the design of new textile fabrics

Abstract In this paper, we propose an Intelligent Decision Support System (IDSS) for the design of new textile fabrics. The IDSS uses predictive analytics to estimate fabric properties (e.g., elasticity) and composition values (% cotton) then prescriptive techniques optimize inputs that feed models types yarns used). Using thousands data records from a Portuguese company, compared two distinct Machine Learning (ML) approaches: Single-Target Regression (STR), via Automated ML (AutoML) tool, Multi-target Regression, deep learning Artificial Neural Network. For analytics, Evolutionary Multi-objective Optimization (EMO) methods (NSGA-II R-NSGA-II) when optimizing 100 fabrics, aiming simultaneously minimize physical property error distance optimized with learned input space. EMO were applied Overall, STR approach provided best results both prediction tasks, Normalized Mean Absolute Error range 4% (weft 11% (pilling) in terms classification accuracy 87% adopting small tolerance 0.01 predicting percentages six fibers cotton). As results, they favored R-NSGA-II method, which tends select Pareto curves are associated average 16% distance.