International Journal of Scientific Research and Engineering Development

Stop relying on predictive machine learning models that find correlations in historical data. Use causal models, like econometric models. Short-term demand forecasting systems fail to distinguish between actual cause and no more than coincidence, which can lead to inadequate pricing measures and a widely known downward price spiral. LLM-DPECI is proposed in this paper, which is an integrated framework which uses causal machine learning, Bayesian Optimization, and retrieval-augmented generation to deliver accurate price decisions which are causally ground and also provide human-readable reports. This study proposes a method that employs the DoWhy and CausalML framework to estimate heterogeneous price elasticities using double ML while explicitly controlling for the confounders, such as competitor pricing, seasonal events, and weather effects. A module for Bayesian Optimization establishes a pseudo-price space using a Gaussian process surrogate to optimize expected revenue under margin constraints. A large language model (GPT/Claude) based RAG pipeline generates easy to understand natural language driven pricing report from complex model outputs for technical actions by non-technical manager and makes them act based on causal insights. Experiments on the Instacart and Walmart M5 public datasets indicate that LLM-DPECI significantly outperforms correlation-based baselines in terms of off-policy demand estimation accuracy and revenue uplift. Users also rated the reports produced by LLM as highly interpretable.This work highlights causal inference as the critical next step beyond predictive ML in commercial pricing systems.

Sundeep Kumar P, Harish Patil, Pratibha S, Sai Shivani K, Anita Patil,"LLM-Powered Dynamic Pricing Engine with Causal Inference for Retail and E-Commerce" International Journal of Scientific Research and Engineering Development, V9(3): Page(147-154) May-June 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.