Abstract :

In The Age Of Data-intensive Enterprise Systems, It Is Essential To Have Accurate And Interpretable Forecasting Of Temporal Data At A Large Scale To Optimize Operations And Strategic Planning. This Study Introduces A Scalable And End-to-end Intelligent Forecasting System, In Which Time Series Regression And Models Are Combined With A Multi-agent Orchestration Layer Implemented On A Large Language Model (LLM). The System Can Handle More Than 15 Million Transactional Records In A MySQL Database, Driven By An Automated Extract, Transform, Load (ETL) And Feature Engineering Pipeline That Produces Lag Features, Rolling Statistics, Seasonal Encodings, And Exogenous Variables. A Hybrid Modeling Approach, Which Uses ARIMA, XGBoost, LSTM And Transformer Models, Allows The Scalable Modeling Of Linear, Nonlinear And Long-range Temporal Relationships. The Experimental Results Indicate That The Transformer Model Is The Most Accurate In Forecasting (mean Absolute Percentage Error [MAPE] = 5.48%), Being At Least 38.5 Times Better Than The Classical Statistical Baselines. The Primary Contribution Of This Work Is The Incorporation Of An Intelligent Agent Based On An LLM With The Ability To Transform Natural Language Queries Into Validated SQL Statements And Model Inference Calls And Produce Explanations In A Human-readable Format Of Predictions. A Practical Analysis Has Verified That, Although The Implementation Of The Longshort-term Memory (LSTM) Neural Network Brings About A Small Amount Of Latency Overhead, Usability And Shortened Decision-making Time By 73% Are Greatly Improved.

Published:

16-6-2026

Issue:

Vol. 26 No. 6 (2026)

Page Nos:

1120-1131

Section:

Articles

License:

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

How to Cite