Prediction Accuracy Increase of Recommender System in Data Scarcity

Recommender system is used to automatically recommend appropriate goods or items to customers. This system analyzes the customer’s interests patterns, predicts the customer’s preference of goods or items and recommends the goods according to the results, in order to find out the relationship of customer and items, and customer’s preference information to the goods or items. This study is conducted to improve the precision of low prediction of customer preferences owing to the scarcity of data in collaborative filtering recommender system. To propose the dilute method of scarcity, that is, preference mean of ratings which individual customer evaluates, preference mean of ratings evaluated in individual item, the overall mean of the means of each user and item are provided and the results are evaluated in each approach. Also, in statistic, the approach using mode as well as mean is provided. In the approach using mode, mode of ratings which each user evaluates, mode of ratings evaluated in individual item, the overall mode of ratings are provided and the results are evaluated in each approach. Finally, the scarcity dilute method which precision of preference prediction is the most superiority is selected, and superior prediction algorithm combining with significance weight is the best method.