# Predicting Future Sales of Retail Products using Machine Learning

**Authors:** Devendra Swami, Alay Dilipbhai Shah, Subhrajeet K B Ray  
**arXiv:** [2008.07779](https://arxiv.org/abs/2008.07779)  
**Date:** 2020-08-18  
**Summary page:** https://www.summarizepaper.com/en/paper/2008.07779v1/

## Key points

- Authors: Devendra Swami, Alay Dilipbhai Shah, Subhrajeet K B Ray
- Study focus: Application of machine learning techniques to forecast future sales using present and past data
- Dataset: Daily sales data from 1C Company, a large Russian software firm
- Objective: Predict total sales for each product and store in the upcoming month using historical data
- Approaches used: eXtreme Gradient Boosting (XGBoost) and Long Short Term Memory (LSTM)
- Evaluation method: Root Mean Squared Error (RMSE) between actual and predicted values
- Findings: XGBoost outperforms LSTM due to higher sparsity in the dataset
- Significance: Highlights the importance of leveraging machine learning models for predictive analytics in retail settings
- Benefits: Enhanced forecasting accuracy, informed decisions on inventory management and marketing strategies

## Summary

In their paper titled "Predicting Future Sales of Retail Products using Machine Learning," authors Devendra Swami, Alay Dilipbhai Shah, and Subhrajeet K B Ray explore the application of machine learning techniques to forecast future sales based on present and past data. The study focuses on a challenging time-series dataset comprising daily sales data provided by 1C Company, one of the largest Russian software firms. The objective is to predict total sales for each product and store in the upcoming month using historical data. To achieve this goal, the authors employ two distinct approaches: eXtreme Gradient Boosting (XGBoost) and Long Short Term Memory (LSTM) based network architecture for learning tasks. They evaluate the performance of these algorithms by calculating the Root Mean Squared Error (RMSE) between actual and predicted target values. Through their analysis, they find that XGBoost outperforms LSTM on this dataset due to its higher sparsity. The study sheds light on the significance of leveraging machine learning models for predictive analytics in retail settings. By effectively utilizing advanced algorithms like XGBoost, researchers and practitioners can enhance forecasting accuracy and make informed decisions regarding inventory management, marketing strategies, and overall business operations. This research contributes valuable insights to the field of retail analytics and underscores the potential benefits of incorporating machine learning into sales prediction processes.