Recommendation System from Microsoft News Data using TF-IDF and Cosine Similarity Methods

Abstract

The rapidly growing information causes information overload, so news portals publish information massively. Readers need time to search and read more news, but the time relevance of news wears off quickly. A recommendation system is needed that can recommend news according to the preferences of readers. This study recommends news using the TF-IDF method. TF-IDF gives weight to each word in the news title, and then looks for similarity between stories using cosine similarity. To prove the accuracy of whether the system recommendation results were actually clicked by the reader, the recommendation results were matched with the reader's news history on the online news portal Microsoft News using a hit-rate. The hit-rate result in this study was 80.77%.