Commercial Intelligence AI-Based Pellet Price Prediction -Digital Transformation in Mining and Metals

Authors

  • Venugopalam Medicherla
  • Aasheesh Raizada

Keywords:

Digital Transformation, Mining Industry, Artificial Intelligence, Machine Learning, Large Language Models, Pellet Price Forecasting, SAP ERP, Commercial Intelligence.

Abstract

Mining & Metals is undergoing a paradigm shift driven by Industry 4.0 technology, increasingly applies artificial intelligence (AI), machine learning (ML) and large language models (LLMs) to commercial decisions, yet pellet pricing in India's iron and steel chain often still rests on manual judgement and spreadsheets. This study develops an AI-based framework to forecast the daily pellet sales price from SAP S/4HANA records, Steel-Mint quotations, competitor indices and plant costing data of an operating producer. Fourteen control predictors—with Sponge iron price (24.7%), Competitor Index 1 (17.3%), and Coal price (13.5%) carrying the largest linear-model weights—were standardised with training-period statistics. An interpretable Linear Regression explained with SHAP and a 200-tree Random Forest were trained on 1,127 daily records from December 2022 to December 2025 and tested on 284 later records. A locally hosted LLM (Ollama 70B) with retrieval-augmented generation (RAG) converts news and regulatory text into a sentiment signal. Error was measured by MAE, RMSE and MAPE and compared with a persistence benchmark. On 284 held-out daily records (6 January to 17 October 2026), the model reached an MAE of ₹578.8/MT and a MAPE of 5.28% (accuracy, 100 − MAPE: 94.72%); 60.9% of forecasts fell within 5% of the observed price. One-day persistence scored 0.41% MAPE, but its error exceeded the model's beyond a horizon of about 30 days. Errors rose sharply in September and October 2026 due to unexpected market volatility and policy shifts. The study documents an explainable pipeline that combines enterprise data, market feeds, ML models and LLM-derived signals, and shows that its value over simple benchmarks depends on forecast horizon, while simultaneously supporting ESG-aligned operational transparency and revenue optimization. The evidence comes from one producer and one commodity.

Downloads

Published

2026-11-16

How to Cite

Medicherla, V., & Raizada, A. (2026). Commercial Intelligence AI-Based Pellet Price Prediction -Digital Transformation in Mining and Metals. Adolescência E Saúde, 1010–1034. Retrieved from https://adolescenciaesaude.com/index.php/aes/article/view/3453

Issue

Section

Original Articles