The objective of this research is to design and develop a tool to evaluate tourists' satisfaction with the attractions recommendation system in Bangkok, Thailand. We have four main stages for the tourist attraction recommendation system. The first stage is to fill imputed missing values with association rules and multiple imputations. The second stage is constructing the tourist attractions recommendation model by ranking the tourist attractions with a ranking method and similarity measurements based on a personal recommender system with cosine algorithm. The third stage is to design and develop the personal recommender website. And the last stage is to evaluate the personal recommender system with four measurements: accuracy, precision, f-measure, and g-mean. The experiment results from a sampling of thirty people found that the tourist attraction recommendation system can: 1) make a positive recommendation 340 times, but 105 times will not meet the needs, and 2) make a negative recommendation 708 times, but 77 times will meet the needs. The results show that the tourist attractions recommendation system has satisfactory performance and reliability with high accuracy, precision, and f-measure, and g-mean values of 85.20%, 76.40%, 78.89%, and 84.26%, respectively. In addition, it was found that the users' satisfaction towards the system was at a high level with a value of 4.60. This means that the proposed tourist attractions recommendation system can be used to recommend personal preferences as well.