A Hybrid Graph-Structure-Based Webpage Recommendation Algorithm based on Weighted Association Rules

Web personalization refers to a set of operations that modify the web experience for a single or a group of users by providing dynamic recommendations based on their behavior patterns. One of the challenges of improving the performance of web personalization algorithms is the simultaneous use of structural data and user browsing information. In this paper, after introducing a weighing criterion, we use a new algorithm based on the graph structure between web pages to offer pages to users. The proposed algorithm combines the link graph structure and the collection of heavy item sets and generates new association rules based on weighted items, or the so-called weighted association rules. In this algorithm, the pages are weighted based on a new measure that reflects the interest of users and the significance of the page. Webpage recommendation is performed based on heavy data. More specifically, the transaction data set is analyzed using a frequent-pattern (FP) tree algorithm, heavy item sets are generated separately, the purity and connectivity of interacting neighbor and non-neighbor items are processed by the Extended Valency algorithm, and finally, the association rules are produced for webpage recommendation. The proposed method for webpage recommendation is called the Extended Heavy Valency algorithm. In the proposed method, the web structure is analyzed using the principles of graph theory and after examining the modes of the graph structure, the weights are calculated based on the concepts of connectivity and purity of neighbor and non-neighbor items at the nodes (links). The proposed weighting formula allows us to calculate the value of a link or node with a relatively high precision. The exact value of the proposed algorithm for the minimum backup value is between 5 to 20, 70 to 82 percent.