Hybrid Recommender System Using Singular Value Decomposition and Support Vector Machine in Bali Tourism

Abstract

When going to make a visit to a tourist area, tourists must determine the place they want to visit. Meanwhile, the desired place has several categories and types. The many types of tourist attractions make tourists confused in determining their choice. Therefore, we focus on developing a hybrid recommendation system by combining several recommendations approaches, namely collaborative filtering, content-based filtering, and demographic filtering. This recommendation system was built to solve the cold start problem that often appears in collaborative filtering and content-based filtering. In this study, weighted and switching techniques were chosen as the hybridization method. These two techniques are used to overcome the weaknesses of each technique so that it becomes a better recommendation system. The singular value decomposition (SVD) algorithm was chosen to be used in collaborative filtering, meanwhile, content-based filtering uses the calculation of cosine similarity values , and demographic filtering uses the support vector machine (SVM) algorithm. The data used in this study is data on tourist destinations in the Bali area obtained from crawling on the TripAdvisor site. In this study, the root mean square error (RMSE) and mean absolute error (MAE) was used to measure the accuracy of the resulting rating prediction. The results of the experiments carried out show that the hybrid method that was built produces better accuracy prediction results than when run separately with an average mean absolute error (MAE) of 0.6660 and a root mean square error (RMSE) of 0.8644.