Web-Based Recommendation System for Dessert Spots in the Renon Area Using the Content-Based Filtering Method

Abstract

The rapid development of information technology has encouraged the application of recommendation systems to help users select culinary places according to their personal preferences. This study aims to design and build a web-based dessert place recommendation system in the Renon area, Denpasar, using the Content-Based Filtering (CBF) method. A total of 61 dessert place data were collected through Google Maps scraping using Apify, covering price, location, and category attributes. The category attribute was transformed using One-Hot Encoding, the price attribute was normalized using Min-Max Normalization, and the distance between the user and dessert places was calculated using the Haversine formula. All attributes were represented as feature vectors, and the similarity between user preferences and dessert place data was calculated using Cosine Similarity to generate the most suitable recommendations. The system was developed using the Laravel framework with the Waterfall method and tested using the Blackbox Testing method. System performance was evaluated using a confusion matrix with 70:30 and 60:40 data-split schemes. The 70:30 scheme testing produced an accuracy of 90.00%, precision of 50.00%, recall of 50.00%, and F1-Score of 50.00%, while the 60:40 scheme produced an accuracy of 92.59%, precision of 75.00%, recall of 75.00%, and F1-Score of 75.00%. The relatively low precision, recall, and F1-Score values were caused by the very small number of truly relevant items in the test data, so a single misclassification had a large impact on these percentages, while accuracy remained high because it was dominated by clearly irrelevant items. These results indicate that the Content-Based Filtering method is reasonably able to provide relevant dessert place recommendations according to user preferences.