Scholarlink

This research undertakes an extensive comparative analysis of sales forecasting methods based on different algorithms on two artificial datasets—Bike Sales Data and Retail Store Inventory Forecasting Data. The study utilizes Regression Analysis and Time Series Analysis (ARIMA, SARIMA, and Holt-Winters) to analyze sales patterns, determine significant trends, and measure predictive accuracy across industries. Preprocessing stages of log transformation and data standardization are applied to overcome data inconsistency and skewness. Model performance is measured by statistical measures like RMSE, MAE, MAPE, and R2. Results show that standard regression models have prominent predicting constraints in dealing with large, unstable datasets, while ARIMA shows relatively superior forecasting ability compared to SARIMA and Holt-Winters. Again, none of the models accurately capture sales variation or exhibit high predictive abilities, emphasizing the need for sophisticated machine learning methods in large-scale sales forecasting problems. The study acts as a core analysis for upcoming studies in embracing more advanced predictive analytics for distinct sales data sets.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Comparative Analysis of Regression and Time Series Forecasting Techniques on Sales Datasets

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Pranjal Mehta,
  • Krishna Singh Rawat,
  • Garv Sehgal,
  • Aditya Sreekumar,
  • Prateek Singhal

摘要

This research undertakes an extensive comparative analysis of sales forecasting methods based on different algorithms on two artificial datasets—Bike Sales Data and Retail Store Inventory Forecasting Data. The study utilizes Regression Analysis and Time Series Analysis (ARIMA, SARIMA, and Holt-Winters) to analyze sales patterns, determine significant trends, and measure predictive accuracy across industries. Preprocessing stages of log transformation and data standardization are applied to overcome data inconsistency and skewness. Model performance is measured by statistical measures like RMSE, MAE, MAPE, and R2. Results show that standard regression models have prominent predicting constraints in dealing with large, unstable datasets, while ARIMA shows relatively superior forecasting ability compared to SARIMA and Holt-Winters. Again, none of the models accurately capture sales variation or exhibit high predictive abilities, emphasizing the need for sophisticated machine learning methods in large-scale sales forecasting problems. The study acts as a core analysis for upcoming studies in embracing more advanced predictive analytics for distinct sales data sets.