TY - JOUR
T1 - A novel hotel recommender system incorporating review sentiment and contextual information
AU - Hu, Ya Han
AU - Tsai, Chih Fong
AU - Sun, Yu Chen
N1 - Publisher Copyright:
© The Author(s), under exclusive licence to Springer Nature Switzerland AG 2024.
PY - 2024
Y1 - 2024
N2 - Hotel recommender systems are crucial tools that assist tourists in filtering through extraneous hotel information, ensuring they find accommodations that align with their preferences. These systems predominantly employ content-based filtering (CBF) and collaborative filtering (CF), including user-based CF (UCF) and item-based CF (ICF), as their core methodologies. Additionally, context-aware recommender systems (CARS) enhance the quality of recommendations by integrating relevant contextual data into the decision-making process. Despite their efficacy, a significant challenge these systems encounter is data sparsity within the user–item matrix, commonly referred to as the traveler–hotel matrix. The sparsity arises as not all users rate every hotel, and some may not provide any ratings at all, resulting in numerous missing ratings and insufficient data for accurately predicting user preferences. This study introduces a novel hybrid hotel recommender system that integrates CBF, UCF, ICF, and CARS methodologies, utilizing travel types as a form of contextual information. A key innovation of our system is the application of sentiment analysis (SA) techniques to extract sentiment data from hotel reviews. This information is then employed to impute missing ratings in the matrix, addressing the issue of data sparsity. Our experimental analysis, leveraging a dataset from TripAdvisor.com, reveals that our proposed system substantially surpasses traditional UCF approaches and other methods that combine CBF, UCF, ICF, and contextual information, with the exception of SA. Notably, our system demonstrates significantly lower Mean Absolute Error and Root Mean Square Error rates compared to baseline models. This advancement not only mitigates the challenge of data sparsity but also enhances the precision and reliability of hotel recommendations for travelers.
AB - Hotel recommender systems are crucial tools that assist tourists in filtering through extraneous hotel information, ensuring they find accommodations that align with their preferences. These systems predominantly employ content-based filtering (CBF) and collaborative filtering (CF), including user-based CF (UCF) and item-based CF (ICF), as their core methodologies. Additionally, context-aware recommender systems (CARS) enhance the quality of recommendations by integrating relevant contextual data into the decision-making process. Despite their efficacy, a significant challenge these systems encounter is data sparsity within the user–item matrix, commonly referred to as the traveler–hotel matrix. The sparsity arises as not all users rate every hotel, and some may not provide any ratings at all, resulting in numerous missing ratings and insufficient data for accurately predicting user preferences. This study introduces a novel hybrid hotel recommender system that integrates CBF, UCF, ICF, and CARS methodologies, utilizing travel types as a form of contextual information. A key innovation of our system is the application of sentiment analysis (SA) techniques to extract sentiment data from hotel reviews. This information is then employed to impute missing ratings in the matrix, addressing the issue of data sparsity. Our experimental analysis, leveraging a dataset from TripAdvisor.com, reveals that our proposed system substantially surpasses traditional UCF approaches and other methods that combine CBF, UCF, ICF, and contextual information, with the exception of SA. Notably, our system demonstrates significantly lower Mean Absolute Error and Root Mean Square Error rates compared to baseline models. This advancement not only mitigates the challenge of data sparsity but also enhances the precision and reliability of hotel recommendations for travelers.
KW - Collaborative filtering
KW - Contextual information
KW - Hotel recommendations
KW - Recommender systems
KW - Sentiment analysis
UR - http://www.scopus.com/inward/record.url?scp=85199341402&partnerID=8YFLogxK
U2 - 10.1007/s41060-024-00598-7
DO - 10.1007/s41060-024-00598-7
M3 - 期刊論文
AN - SCOPUS:85199341402
SN - 2364-415X
JO - International Journal of Data Science and Analytics
JF - International Journal of Data Science and Analytics
ER -