Authors: Assistant Professor Dr.N.Indrajith, Assistant Professor Mrs.K.Shanmuga Priya
Abstract: Enterprise-level data availability has provided unprecedented levels of possibility in making decisions through evidence but the ability to convert raw data into intelligence still remains challenging. In this paper, we present a comprehensive framework for AI-enabled predictive analytics that can serve as an effective link between analyzing the data and decision-making process. In particular, our methodology incorporates distributed computing, feature extraction, hybrid predictive modeling based on XGBoost, LSTM, and Transformer neural networks, as well as prescriptive decision-making process optimization. Experiments demonstrate that the hybrid model exhibits excellent predictive performance with R² = 0.972 that significantly exceeds standalone ones. At the same time, the use of prescriptive approach results in 18.7% cost savings and 23.4% increase in decision accuracy.
