| Summary: | The challenge of finding an ideal homestay that aligns with user preferences is significant due to the volume of available choices and the limitations of existing search filters on accommodation websites. Users face several issues when searching for homestays online, such as limited search filters, lack of location or landmark information, irrelevant recommendations, outdated or inaccurate homestay details, and time-consuming search processes. Additionally, websites often mix homestay listings with other types of accommodations, leading to user confusion and inefficient search experiences. The objective of this study is to investigate the requirement needed to develop a homestay recommender system using Content-based filtering and to evaluate the accuracy and performance of the homestay recommender system. The data used in this research, which contain the criteria of the homestays, were collected from the online accommodation website. Recommendation list is generated using Content-based filtering approach, which includes few techniques such as Term Frequency-Inverse Document Frequency (TF-IDF) to analyze the importance of terms in the dataset, Cosine Similarity to measure the similarity between the user's preferences and homestays characteristic and K-Means algorithm for the evaluation part. The system's performance is evaluated using precision, recall, F1-score, Root Mean Square Error (RMSE), and Mean Absolute Error (MAE). The prototype achieved an accuracy of 79.3%, a recall of 93%, and an F1 score of 86%. The RMSE and MAE were recorded at 9.95 and 9.42, respectively. As for future works, expanding the dataset and implementing a proper database to store the user profile could be done to enrich the recommendation process. © 2024 IEEE.
|
|---|