TEPM: Traveler Enrollment Prediction Mechanism using BERT-based Feature Clustering and LSTM Models

doi:10.1504/IJAHUC.2024.138748

淡江大學機構典藏 > 工學院 > 資訊工程學系暨研究所 > 期刊論文 > Item 987654321/126325

Please use this identifier to cite or link to this item: https://tkuir.lib.tku.edu.tw/dspace/handle/987654321/126325

Title:	TEPM: Traveler Enrollment Prediction Mechanism using BERT-based Feature Clustering and LSTM Models
Authors:	Tsai, Chung-You;Su, Ming-Yang;Chuang, Christopher;Chang, Chih-Yung;Roy, Diptendu Sinha
Keywords:	tour group prediction;feature clustering;natural language processing;BERT model;LSTM enrolment prediction
Date:	2024-05-29
Issue Date:	2024-10-02 12:05:37 (UTC+8)
Abstract:	The prediction of whether a tour group will form or not has a significant impact on travellers' future itinerary planning and travel agencies' control over hotel and flight bookings. Traditional methods rely solely on historical data, therefore lacks accuracy due to diverse tour attributes. The proposed mechanism, called TEPM, divides the enrolment prediction into three stages, including clustering, classification and prediction. Firstly, it clusters the tours to several groups according to the enrolment data. Secondly, natural language processing techniques are used to convert tour advertisements into feature documents. The BERT is employed to learn the relationship between advertisement feature documents and clusters. This enables the prediction of the group to which a given tour advertisement belongs. Finally, in the prediction stage, this paper employs dedicated LSTM models for each cluster to predict the number of enrolees. Experiments show that this approach performs well in terms of precision, recall, and F1 score.
Relation:	International Journal of Ad Hoc and Ubiquitous Computing , vol. 46, no. 1, pp. 14-26
DOI:	10.1504/IJAHUC.2024.138748
Appears in Collections:	[資訊工程學系暨研究所] 期刊論文

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	75	View/Open

Loading...