İşbirliğine dayalı filtreleme tavsiye sistemi: Karşılaştırma çalışması

Recommender systems (RS) have been getting serious attention in solving information overload problems by suggesting to users, items that are potentially of interest to them. Recommendation systems usually produce a number of suggestions in one of the given techniques. The RS divided into three types: Content-based filtering, Collaborative Filtering and Hybrid recommender system. Collaborative Filtering (CF) is the most popular recommendation technique and widely adopted in many commercial domains. However, CF does not consider any additional information, making it difficult to solve the cold-start and data sparsity problems. As found in most knowledge, likewise, Recommender system has some problems such as cold-start, data sparsity and scalability and so on; many researches are done to solve these problems and to increase the accuracy of the prediction. This study will produce comparison of different algorithms in RS, which are KNN, SVD and Naïve Bayesian to check the best performance of them. Two different sizes of dataset is applied (80% with 20%) and (60% with 40%) each size includes training and testing. Moreover, the performance is computed by three metrics (MAE, RMSE and time). The results revealed that Naïve Bayesian has highest accuracy in both metrics MAE and RMSE for both sizes of the datasets for; whereas in term of time the result was vary, depending on size of the dataset with the technique of the algorithm.