Multi-Model Demand Forecasting of Stock Keeping Units Quantities

Swastik Agarwal; Harshith Suresh

doi:10.36948/ijfmr.2025.v07i03.48132

Multi-Model Demand Forecasting of Stock Keeping Units Quantities

Author(s)	Swastik Agarwal, Harshith Suresh
Country	India
Abstract	In an increasingly volatile and demand-driven market landscape, accurate demand forecasting is critical for optimizing inventory levels, minimizing operational waste, enhancing customer satisfaction, and maintaining end-to-end supply chain agility. This project proposes a robust and extensible demand forecasting pipeline that combines traditional statistical models with advanced machine learning approaches—specifically ARCH (Autoregressive Conditional Heteroskedasticity), GARCH (Generalized ARCH), Markov models, and Facebook Prophet—to capture diverse temporal patterns including volatility clustering, state transitions, trend, and seasonality. A key innovation lies in its data preprocessing strategy, where missing values are handled through multiple imputation methods such as forward-fill, backward-fill, median substitution, rolling averages, and statistical trimming. Each model is evaluated with these imputation techniques, and the most effective model-imputation pairing is selected based on a comprehensive set of performance metrics: Weighted Absolute Percentage Error (WAPE), Mean Absolute Percentage Error (MAPE), Root Mean Squared Error (RMSE), and the Coefficient of Determination (R²). The dataset, consisting of SKU-level order quantity time series, is split using a train-test framework where the last six months are reserved for testing to mimic real-world forecasting scenarios. This empirical, metric-driven approach enables the selection of the best-performing forecasting strategy, ensuring both accuracy and generalizability. The pipeline is designed to be modular, allowing for easy integration of additional models or imputation strategies, and is applicable across various domains including retail, manufacturing, and logistics. Future directions involve extending the pipeline to support real-time data ingestion, automated feature selection, and deep learning models such as LSTM and Transformer-based architectures for enhanced long-term forecasting accuracy.
Keywords	Demand Forecasting, Time Series Models, Data Imputation, ARCH, GARCH, Markov Models, Prophet, WAPE, MAPE, Forecasting Pipeline
Field	Computer > Artificial Intelligence / Simulation / Virtual Reality
Published In	Volume 7, Issue 3, May-June 2025
Published On	2025-06-15
DOI	https://doi.org/10.36948/ijfmr.2025.v07i03.48132
Short DOI	https://doi.org/g9qqpn

View / Download PDF File

E-ISSN 2582-2160

doi

CrossRef DOI is assigned to each research paper published in our journal.

IJFMR DOI prefix is
10.36948/ijfmr

Downloads

Research Paper Format Copyright Permission Form and Undertaking Form Cover Page Vol 7 Isu 3 Cover Page Vol 7 Isu 2 Cover Page Vol 7 Isu 1

All research papers published on this website are licensed under Creative Commons Attribution-ShareAlike 4.0 International License, and all rights belong to their respective authors/researchers.

CC-BY-SA

About IJFMR Fees & Payment Current Issue Publication Archive	Submit Research Paper Track Submission Status Publication Guidelines Publication Ethics Peer Review & Plagiarism	Join as a Reviewer Editors & Reviewers Reviewer Referral Program Get Reviewer Membership Certi.	Website/Journal Policies Usage Policy Content Policies Privacy Policy

Contact Us		+91-9687-828-838	editor@ijfmr.com

International Journal For Multidisciplinary Research

A Widely Indexed Open Access Peer Reviewed Multidisciplinary Bi-monthly Scholarly International Journal

Multi-Model Demand Forecasting of Stock Keeping Units Quantities

Share this