This work develops a novel approach based on Machine Learning (ML)-assisted Quality Function Deployment (QFD) to sift the gold from the stone. It includes Time-Varying Filter-based Empirical Mode Decomposition (TVF-EMD), Deep Ensemble Random Vector Functional Link (DE-RVFL), and a Bayesian optimization algorithm for optimizing the shaped DE-RVFLTVF-EMD hyperparameters. This approach makes it possible for the proposed methods to be dynamic enough to deal with the data's volatility, complexity, uncertainty, and ambiguity. It is demonstrated that incorporating TVF-EMD to provide time-frequency analysis along DE-RVFL, and goal-oriented social media analytics boosts the performance of out-of-sample predictions statistically and compensates for the “warranty data maturation” effect. The proposed algorithm's Root Mean Square Error (RMSE) decreases by 23.37%-88.76% relative to other benchmark cutting-edge models. This study contributes significantly to the services management com