Web-Based Smartphone Recommendation System Using TF-IDF and Cosine Similarity Methods

Abstract

The advancement of the digital era urges the use of smartphones as an essential instrument for society. However, the large variety of types and features actually triggers a dilemma for ordinary buyers who want to find a gadget according to their personal preferences. The average conventional search engine still relies on static filters that are less responsive to human language queries. Therefore, this study initiated a website-based gadget recommendation platform that adopts the Term Frequency-Inverse Document Frequency (TF-IDF) method combined with Cosine Similarity. This approach is executed so that the system can extract meaning from user-free text and calculate its proximity to the phone's specification database. With the development method (prototype), the system architecture is built using PHP, Laravel framework, and MySQL. The processed data is sourced from trusted sites in the 2024-2026 period. The output of this research is in the form of an intelligent application that is able to sort the most relevant products along with direct comparison facilities. The results of the Cosine Similarity calculation show that the system is able to provide recommendations precisely, as evidenced by testing 20 smartphone product data with the query "hp 6 million gaming" where  the highest similarity  value was calculated to reach a score of 0.381 on the Poco F7 product, which makes it the most accurate recommendation to support the final decision of consumers.