Please use this identifier to cite or link to this item: https://ir.swu.ac.th/jspui/handle/123456789/29414
ชื่อเรื่อง: Machine Learning Approach to Predict E-commerce Customer Satisfaction Score
ผู้แต่ง: Wangkiat P.
Polprasert C.
Keywords: classification
customer satisfaction
e-commerce
machine learning
rating prediction
วันที่เผยแพร่: 2023
สำนักพิมพ์: Institute of Electrical and Electronics Engineers Inc.
บทคัดย่อ: In this paper, we investigate the performance of machine learning (ML) to predict customers' satisfaction score from the sales dataset collected by Olist, the Brazilian e-commerce company. Customer satisfaction score is categorized into 4 classes: Low, Average, Good and Excellent where majority of sales orders receive Excellent score. Inspired by the fact that delivery duration and product rating score obtained from other customers' purchase are one of the main factors that influence customer's satisfaction score, we propose a feature engineering method that creates delivery duration and the average product rating score which are used as the main features in the ML model. We employ Random Forest (RF), Logistic Regression (LR), and K-Nearest Neighbor (K-NN) to predict customers' satisfaction score and their performance are compared with the baseline model which predicts the customer satisfaction score using the average product rating score. Results show that RF model yields the best performance with the average precision, recall, and macro F1 equal to 0.34, 0.36, and 0.32, respectively. In addition, RF achieves the best recall equal to 0.43, 0.33 and 0.33 for Low, Average and Good classes. The mean and SD of the product rating are two features with the highest feature importance equal to 0.313 and 0.087, respectively. © 2023 IEEE.
URI: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85164282311&doi=10.1109%2fICBIR57571.2023.10147542&partnerID=40&md5=f0a71c546c8a44bd779d2f23f942971a
https://ir.swu.ac.th/jspui/handle/123456789/29414
Appears in Collections:Scopus 2023

Files in This Item:
There are no files associated with this item.


Items in SWU repository are protected by copyright, with all rights reserved, unless otherwise indicated.