A Retail Product Prediction Model Incorporating Machine Learning and Propensity Score Matching Method

Existing machine learning methods for retail product sales forecasting often rely on their own time series data and tend to ignore the correlation between the target retail product and other products. In this paper, we take cigarette product retail sales prediction as an example, and use the neighbo...

Full description

Saved in:
Bibliographic Details
Published in:2024 7th International Conference on Computer Information Science and Application Technology (CISAT) pp. 1030 - 1038
Main Authors: Chen, Kaidi, Liang, Xuexia, Mo, Yuhua, Li, Yuankun, Xu, Liangben, Kong, Weili
Format: Conference Proceeding
Language:English
Published: IEEE 12.07.2024
Subjects:
Alcoholic beverages
Analytical models
Cigarette Release Strategy
Clustering algorithms
component
Data models
Decision making
Decision-making Quality
Heuristic algorithms
KMeans++ Clustering Algorithm
Machine learning
Machine learning algorithms
Machine Learning XGBoost Algorithm
Prediction algorithms
Predictive models
Propensity Score Matching
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Existing machine learning methods for retail product sales forecasting often rely on their own time series data and tend to ignore the correlation between the target retail product and other products. In this paper, we take cigarette product retail sales prediction as an example, and use the neighboring related alcohol sales data to predict cigarette product sales through the propensity score matching (PSM) method, KMeans++ clustering algorithm and XGBoost algorithm. Specifically, we use cigarette and alcohol sales data from Guangzhou, China for the years 2021-2022 as the study sample, and the results show 1) Compared with using only tobacco data and machine learning algorithms, the modeling accuracy improves from 82.79% to 91.34% by introducing neighboring alcohol sales data and combining PSM and machine learning XGBoost algorithms. 2) The machine learning XGBoost algorithm identifies successive important features of cigarette sales data: category, specification, price range, and revenue from neighboring alcohol sales (new features generated by the PSM and KMeans++ algorithms). The conclusions show that the PSM algorithm and the machine learning XGBoost algorithm can incorporate other retail products to predict target products and provide an empirical application for predicting retail sales based on associations with neighboring related products
DOI:10.1109/CISAT62382.2024.10695432