Using artificial intelligence models to predict consumer consumption habits trends
Zhang, Qixiao (2025)
2025
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-202504096024
https://urn.fi/URN:NBN:fi:amk-202504096024
Tiivistelmä
With the rapid development of e-commerce, using data-driven prediction technology to optimize product sales and user experience has become a hot topic in the industry. Based on Amazon sales data, this paper carried out the prediction analysis of product sales and user behavior (D’Agostino, 2018, p. 10). Through in-depth mining of key features such as product categories, discount ratios, user ratings and reviews in the data, this paper constructs a variety of machine learning and deep learning models to accurately predict sales trends and user behavior.
In data preprocessing, missing value imputation, outlier detection and feature engineering are used to improve the practical ability and prediction accuracy of the model.
Experimental results show that the prediction model based on XGBoost and random forest performs well in handling complex interactions and nonlinear relationships. (Brownlee J, 2021). Among them, XGBoost excels in iterative boosting and adaptively managing varied data distributions, while random forest remains stable when dealing with high-dimensional features. (Cornell University, 2016) Through feature importance analysis, the impact of discount strategies and user ratings on sales trends is further examined (Weyders, 2021, p. 5).
This study provides an important reference for optimizing the pricing strategy, inventory management and recommendation system design of e-commerce platform, and provides a practical exploration direction for the research of prediction model in the field of e-commerce.
In addition, this study can also provide data support for the development of marketing strategies, help merchants more accurately target users and improve sales conversion rates.
In data preprocessing, missing value imputation, outlier detection and feature engineering are used to improve the practical ability and prediction accuracy of the model.
Experimental results show that the prediction model based on XGBoost and random forest performs well in handling complex interactions and nonlinear relationships. (Brownlee J, 2021). Among them, XGBoost excels in iterative boosting and adaptively managing varied data distributions, while random forest remains stable when dealing with high-dimensional features. (Cornell University, 2016) Through feature importance analysis, the impact of discount strategies and user ratings on sales trends is further examined (Weyders, 2021, p. 5).
This study provides an important reference for optimizing the pricing strategy, inventory management and recommendation system design of e-commerce platform, and provides a practical exploration direction for the research of prediction model in the field of e-commerce.
In addition, this study can also provide data support for the development of marketing strategies, help merchants more accurately target users and improve sales conversion rates.