An integrated predictive-optimization framework for adaptive decision-making: model and algorithm design

Abstract

In response to the complex decision-making requirements during the process of business digital transformation, this paper constructs an AI-driven intelligent decision-making model and a path framework. Based on the "driver-constraint" quantitative model and stage efficiency indicators S and Etotal, the transformation motivation, constraints, and path evolution are parameterized and characterized; further, around multisource heterogeneous business data, feature engineering and standardized preprocessing procedures are designed, a regression prediction model with L2 regularization is established, and a cost-time multiobjective optimization model is constructed by combining genetic algorithms, forming a "prediction-optimization-execution" closed-loop decision-making mechanism; at the same time, through decision process modeling, the system integration of the data layer, model layer and business logic is achieved. Experiments show that compared with rule engines and traditional regression models, the intelligent decision-making model achieves accuracy of 0.92, recall rate of 0.89 and F1 score of 0.90 in terms of precision, recall rate and F1 score, and the prediction error decreases from 12.5% to 3.5% in continuous 10-day tests, verifying the stability and engineering application value of the proposed model in complex business scenarios.