Research On Knowledge Recommendation Based On User Interest Topic Model

In the era of knowledge economy,knowledge has become an important factor driving organizational development.With the expansion of the size of the organization and the development of informatization,more and more knowledge is accumulated,the knowledge base is exploding,and people are caught in the dilemma of "knowledge overload" and "knowledge trek".Knowledge users are looking for Knowledge will take more time.Therefore,alleviating the obstacles to knowledge acquisition and dissemination,improving the efficiency of knowledge utilization,and bringing convenience to knowledge users has become an urgent problem.Knowledge recommendation can push the right knowledge to the users who need it in the right way at the right time,which is an important way to alleviate this problem.Collaborative filtering algorithm is widely used in knowledge recommendation because of its simple characteristics.However,the algorithm has two shortcomings: in terms of recommendation accuracy,the algorithm only uses the user's historical behavior as the source of interest,ignoring knowledge semantics.In terms of algorithm efficiency,the algorithm needs to traverse the entire user list to get user neighbors.This will reduce the efficiency of the algorithm in the case of increasing user scale and knowledge scale in the recommendation system.Here,based on user-based collaborative filtering,this paper proposes a knowledge recommendation algorithm based on user interest topics and user clustering.First,the LDA model is introduced to mine user interest topics.As we all know,as a cohesive and systematic information content,what can guide people's thoughts and behaviors is its semantic connotation.Knowledge semantics can more accurately express user interests and improve the accuracy of recommendations,which is crucial in knowledge recommendation.This article uses the LDA model to perform topic mining on user knowledge text,and expresses user interest in the form of topics and their probability distributions.Aiming at the shortcomings of the number of topics in the LDA model,which is difficult to determine and does not reflect dynamics,this paper proposes an LDA topic model that optimizes the number of topics and adds a time factor,so that the model can more accurately acquire knowledge topics and express user interests more clearly and dynamically..Secondly,FCM algorithm is introduced to cluster users to shorten the search range during similarity calculation.Collaborative filtering algorithms have problems of sparse data and low efficiency.This paper performs user clustering before recommending.Aiming at the deficiency of FCM algorithm in user clustering,this paper uses Wasserstein distance instead of Euclidean distance.After the distance measurement method changes,the solution of the cluster center will change accordingly.This paper uses artificial immune algorithm to find new cluster centers.Then,the knowledge recommendation index is calculated according to the collaborative filtering algorithm formula,and a knowledge recommendation list is constructed.Finally,the progress and effectiveness of the algorithm are verified through data collection and experimental analysis.