Research On Recommendation Algorithm Based On Hidden Markov Model

Under the background of information age,the concept of big data is rising.Massive data has appeared in all fields of the world,and the world has entered the era of data explosion.The total amount of these data is far beyond the range that ordinary people can accept,process and effectively use.The massive information is full of redundant and lack of value information,which further interferes with people's screening and utilization of information,resulting in information overload.The recommendation algorithm is a practical method to deal with the problem of information overload.Most of the existing recommendation algorithms are based on the user's interests and preferences will not change with time.But this is a rather bad assumption.In reality,everyone's interest will always change.It is inevitable to be biased to solve problems from a static point of view.The hidden Markov model has a strong ability of large time series modeling,so this paper uses this feature to design a recommendation algorithm based on this model.The focus of this method is to overcome the dynamic change of user preference in recommendation algorithm.This paper combines the advantages and disadvantages of collaborative filtering recommendation algorithm and probability model recommendation algorithm,and designs a recommendation algorithm based on Hidden Markov model.This recommendation algorithm not only solves the dynamic change of user preferences,but also provides a powerful information mining ability for the system.At the same time,it avoids the problem that the unknown user preferences cannot be mined in the probability model.If the hidden Markov model is directly introduced to the recommendation algorithm,it will face a series of problems.Therefore,this paper puts forward corresponding improvement measures.It includes the dynamic expansion method of user training data set and the improved model training method.These contents avoid some problems in the hidden Markov model,so that this research idea can effectively solve the problem.Finally,a series of comparative experiments are designed on the distributed big data computing platform to verify the proposed algorithm.Finally,the experimental results show that the proposed algorithm based on Hidden Markov model is better than some classical algorithms.The accuracy,coverage and f index are all improved.