Analysis of Demand Forecasting Trends using Hybrid Regression Model in Comparison with Seasonal Autoregressive Integrated Moving Average with eXogenous Factors Model
Adibhatla Ajay Bharadwaj, Muthumanickam Gunasekaran · 2024
The main goal of this research is to improve the precision of predicting product demand trends through the application of the Novel Hybrid Regression (NHR) machine learning model. This approach will be compared with the Seasonal Autoregressive Integrated Moving Average With eXogenous Factors model (SARIMAX). To achieve this objective, a store item demand forecasting dataset will be utilized. The data will be initially segmented and split into two groups, each consisting of a sample size of 20 or more, to facilitate a thorough comparative analysis. The ultimate sample size is established through a G-power pretest, configured with parameters set at 80%, level of significance 0.05%, and CI 95%. Post-model evaluation, the findings reveal that the Novel Hybrid Regression model outperformed, achieving an accuracy of 84.16%, surpassing the SARIMAX model, which exhibited an accuracy of 72.61 %. The statistical analysis, employing an independent sample t-test, reveals a significant contrast between the two models with a p-value of 0.000$(\mathrm{p} < 0.05)$in a two-tailed test, indicating a substantial difference. In conclusion, the findings of this research strongly support the proposition that the Novel Hybrid Regression model provides a more accurate prediction of demand trends for the product compared to the SARIMAX model. This has practical implications for improving the precision of demand forecasting in various industries.