การประเมินความพึงพอใจของผู้ส่งในบริการขนส่งพัสดุด้วยการวิเคราะห์ความรู้สึกจากข้อความ

Abstract

The research aimed to develop a model to assess the satisfaction of the users of parcel services. The assessment applied the sentiment analysis on textual data, using feedback from the users of parcel services on www.phatsadu.com. The research compared the performance of each sentiment analysis model: Thai Bidirectional Encoder Representations from Transformers (Thai BERT) which included 1) wangchanberta-base-att-spm-uncased; 2) wangchanberta-base-wiki-spm; and 3) wangchanberta-base-wiki-sefr and machine learning tools which included 1) Logistic Regression; 2) Support Vector Machines; and 3) Random Forest Classifier. The purpose was to have the models categorize sentiments into groupings which the researcher had determined based on the feedback of the users of parcel services. The findings revealed that Logistic Regression showed the best performance for Thai texts with 91% accuracy. Furthermore, the most frequently mentioned topics by the users of parcel services were delivery time (41.19%), service (36.39%), internal system (13.05%), staff (5.18%), security (3.21%), and pricing (0.98%). Then, the Analytic Hierarchy Process (AHP) was performed to assess the satisfaction among the users of parcel services of the top four companies in Thailand. The results indicated that, based on various data-driven criteria, the company with the highest score was Thailand Post, followed by Kerry Express Thailand , Flash Express, and J&T Express, respectively.