Towards On Personalization Process Recommendation Based On Hybrid Collaborative Filtering

Traditional process recommendation mainly provides workers recommendationwhen modeling a process. It helps speed up the business process modeling and reduceerrors that may occur during developing process at the same time. However, it doesnot take into account the user personalized needs and then lack of flexibility.Therefore, aiming at the personalized requirements of users as well as the flexibilityof process, we propose user-oriented personalization process recommendationtechnique. This method not only can help users accelerate completing process, butalso to meet the individual needs of users and increase the flexibility of businessprocesses. Research on personalization process recommendation technology, hasimportant research significance and application value.In this thesis, we have research on process recommendation and personalizationrecommendation. Having read a lot of literatures, we combines processrecommendation and personalization recommendation, and then implement processnodes recommended according to users personalized needs.Firstly, introduce the method of obtaining an executable set of candidate activity.This section divides all processes into two parts, the influential upstream path andcandidate activity nodes, to form a process-node mapping table. And throughcalculating the similarity between the parts of process of the current user has toperform and all influential upstream path in the mapping table, finally get thecorresponding candidate node set. This thesis considers three parts of the processnodes, the process structure model and the process behavior as similaritymeasurement. Finally, and the experimental evaluates the effect of mixed processsimilarity measurement.Secondly, introduce user similarity measure method and the improved usersimilarity measure method. In order to form the neighbor users, the most direct way isto calculate the user similarity. This thesis study on the similarity between the userswho use process, the times of users use process indicate user preferences, so wepropose the user similarity measurement method based on user preferences.Then, describe and evaluate the personalization process recommendation methodproposed in this thesis. This part describes how the candidate nodes can be recommended. It does some comparison with FlowRecommeder on effectiveness andefficiency. Experiments show that the personalization process recommendation basedon collaborative filtering method has better effectiveness and efficiency.Finally, we have a summary of work and contributions on the personalizationprocess recommendation, and look forward to the future direction of personalizationprocess recommendation.