Implementation of term frequency-inverse document frequency (TF-IDF) and Word2Vec in traditional medicine recommendation system based on content-based filtering

According to World Health Organization (WHO), traditional medicine is the culmination of all the knowledge, abilities, and practices derived from the theories, beliefs, and experiences that are unique to various cultures and that are used to maintain health as well as to prevent, diagnose, treat, or improve physical and mental illness. recently classified traditional herbal therapy as comprised of medicinal techniques that have existed, frequently for hundreds of years, prior to the establishment of modern medicine. The lack of easily accessible information regarding the description and efficiency of traditional medicine makes it difficult for users to understand the benefits of each type of traditional medicine. Because of this, a recommendation system is needed that aims to facilitate users in finding traditional medicine that suit their preferences. This research proposes a traditional medicine recommendation system with the content-based filtering method using a combination of term frequency-invers document frequency and Word2Vec feature extraction. This method analyzes the traditional medicine description text and recommends based on word weights and semantic relationships between words. Results show optimal performance at dimensions 50–200 and window sizes 9–15 for the combination of term frequency-invers document frequency and Word2Vec, while term frequency-invers document frequency alone reaches 80% of accuracy and Word2Vec has lower performance (4–14%) across a wide range of parameter experiments. Based on optimal result above, this recommendation system can be applied to obtain information of traditional medicine that suitable with people needed by adjust the best model of dimensions and window size