Design and implementation of a user interest recommender algorithm based on improved FP-growth

In this paper, a novel algorithm of mining association rules based on improved FP-growth which focuses on precise user interest recommendation, is proposed for the system without user ratings but a large amount of browsing data. The improvements based on traditional FP-growth can be particularized in the following: First of all, the frequency of user browsing is normalized and multiplied by forgetting function in order to reflect the current interest of users. Besides, dynamic support is adopted to filter out non-interest items and enhance the recommending accuracy. Moreover, activity of item (AOI) is employed as a feasible scheme in the FP-tree to solve the cold start problem. Then, the improved algorithm is verified by the browsing data of the smart contracts in the Enterprise Operation System (EOS) blockchain. The experimental results show that the improved algorithm performs satisfactory in the user interest recommendation.