Optimizing E-Commerce Pricing Strategies: A Comparative Analysis of Machine Learning Models for Predicting Customer Satisfaction

Аннотация

INTRODUCTION transaction records, capturing a diverse array of In the rapidly evolving landscape of e-commerce, variables including customer satisfaction scores, pricing strategies play a crucial role in influencing pricing information, and demographic attributes. customer satisfaction and driving business success. This comprehensive dataset, aggregated from As businesses seek to enhance their multiple e-commerce platforms, provides a robust competitiveness and optimize their pricing foundation for training and evaluating the models. approaches, leveraging advanced data-driven Data preprocessing was a critical phase in this methodologies has become increasingly important. study, involving essential steps such as outlier Machine learning models offer powerful tools for removal, missing value imputation, normalization analyzing and predicting customer behavior, of variables, and encoding of categorical data. enabling businesses to make informed decisions These preprocessing techniques ensured the that align with market dynamics and consumer dataset's quality and suitability for machine expectations. learning applications, enabling accurate and This study explores the application of various reliable model training. machine learning techniques to optimize e- The evaluation of model performance utilized a commerce pricing strategies, focusing on range of metrics—Mean Absolute Error (MAE), improving customer satisfaction through precise Root Mean Square Error (RMSE), R-squared (R²), and data-driven pricing decisions. By evaluating and F1-Score—each offering valuable insights into and comparing the performance of five prominent different aspects of model effectiveness. By machine learning models—Linear Regression, systematically comparing these metrics, the study Decision Trees, Random Forest, Support Vector assesses how well each model manages prediction Machines (SVM), and Neural Networks—the study accuracy, error handling, explanatory power, and aims to identify the most effective approach for balance between precision and recall. predicting pricing outcomes and enhancing customer satisfaction. Ultimately, the study provides actionable recommendations for businesses seeking to The dataset employed for this analysis comprises a enhance their pricing strategies. The comparative rich collection of historical e-commerce analysis highlights the strengths and limitations of https://www.theamericanjournals.com/index.php/tajet 7 THE USA JOURNALS THE AMERICAN JOURNAL OF ENGINEERING AND TECHNOLOGY (ISSN – 2689-0984) VOLUME 06 ISSUE09 each model, offering guidance on selecting the most The introduction of Neural Networks, particularly appropriate approach based on specific business deep learning models, has marked a significant needs, resource constraints, and the importance of shift in pricing optimization. LeCun, Bengio, and model interpretability versus predictive accuracy. Hinton (2015) discuss the advantages of Neural Networks in capturing intricate relationships In summary, this research underscores the within large datasets, which traditional models significance of employing sophisticated machine might miss. Neural Networks, with their ability to learning techniques in optimizing e-commerce learn non-linear patterns, have been shown to pricing strategies. By integrating these techniques outperform other models in various tasks, into pricing decisions, businesses can better align including pricing strategy optimization their pricing strategies with customer (Goodfellow, Bengio, & Courville, 2016). The expectations, improve satisfaction, and achieve application of Neural Networks in e-commerce more favorable business outcomes. pricing, as indicated by Nguyen et al. (2019), has LITERATURE REVIEW led to improvements in prediction accuracy and The application of machine learning (ML) customer satisfaction due to their capability to techniques to e-commerce pricing strategies has handle large and complex datasets. garnered significant interest in recent years. These Evaluating the performance of ML models is crucial methodologies offer advanced analytical for understanding their effectiveness in pricing capabilities that can substantially enhance pricing strategies. Metrics such as Mean Absolute Error decisions and improve customer satisfaction. This (MAE), Root Mean Square Error (RMSE), R-squared literature review provides an overview of key (R²), and F1-Score are commonly used to assess research, and findings related to ML in e-commerce model accuracy and reliability. According to pricing, focusing on the use of various predictive Hyndman and Athanasopoulos (2018), MAE and models and their impact on pricing strategies. RMSE are fundamental for measuring prediction Machine learning models have been extensively errors, with RMSE providing a more sensitive explored for their potential to optimize pricing assessment due to its penalization of larger errors. strategies in e-commerce. Research by Aggarwal R² is useful for un