Tweet hybrid recommendation based on latent dirichlet allocation

Abstract

Recommender system was created to recommend products to users that user may interest. The most recommender systems use two kinds of recommendation techniques which are collaborative filtering (CF) and content-based filtering (CBF). CF use combination of ratings from users in the system who are similar to target user to recommend. Users who are similar to the target user are called neighbors. Therefore, CF will give variety recommendations. CBF uses the past behavior of the target user to find a similar item to the target user’s behavior to recommend. Nowadays, there are many data on social networks including tweet data in the Twitter. Thus, many researchers have studied recommender systems which based on tweet using latent Dirichlet allocation (LDA) to extract latent data from observed data. However, those researches use either CF or CBF with LDA only. However, disadvantages of CF are sparsity and cold-start problem. So, the system cannot efficiently recommend. For CBF, it cannot recommend a new product that users may be interested. Therefore, this research recommends tweets base on hybrid recommender system with LDA, which combines CF and CBF to solve disadvantages of CF and CBF. From experimental results, the proposed method outperforms in term of mean absolute error and coverage.