Research And Application Of Diversity In Recommendation System Based On Item-CF Algorithm

With the increasing rapid growth of the mobile Internet and the increasingly close quality of people's daily life in the information age,the mobile and PC-side information data has grown exponentially.The useful information required by each user due to occupation,location,habits,hobbies,and gender is diverse,and the proportion of massive information is very small,occupying only a small part of it.It is very necessary and meaningful to find data information that is valuable to users in a network with information overload for a limited time.In order to solve the problem that the information is large and the valuable information is overwhelmed,the recommendation algorithm and the search engine are the key ways to solve the overload problem of the user information.The search engine filters,processes,matches the data in the background of the system according to the parameters input by the user,and then returns the data information to the front end to display to the user.When using the search engine to find information,there are two disadvantages.The first one is,Users need to input data actively,otherwise most of them are advertisements or spam.The second is that the data input by the user is inaccurate or cannot provide relevant keywords.In this case,the user's useful information cannot be provided,resulting in very Poor experience and the inability to increase user viscosity.Compared with the search engine,the recommendation algorithm does not require the user to input explicit data,and only needs to actively display the useful information to the customer according to the historical traces of the user,such as: search record,click browsing,adding a shopping cart,placing an order,and evaluating..After long-term information development,the recommendation algorithm has been widely used in all aspects of our life.However,most of the recommended algorithms are mainly focused on the accuracy of recommendation,which makes the recommendation system diversified and the recommendation result is boring.In this paper,the research and application of Item-CF in the recommendation system is mainly done the following work.(1)Explain the reasons and purpose of improving the personalization and diversity of recommendation system by understanding the use status of recommendation algorithms at home and abroad,and demonstrate the necessity of improvement in the current lack of individualization and diversity.(2)Since the contribution of active users and inactive users to the same behavior of products is different,the contribution of inactive users is greater and the data is more realistic.In order to prevent active users from brushing and hot items,it will greatly affect The unfairness caused by the heat of similar items to the recommendation of the goods,through the hot penalty for active users or hot goods,greatly assists the recommendation of new products,and improves the diversity of recommendation algorithms through the similarity of items.The normalization further enhances the diversity of recommendations.(3)When calculating the user's interest in the item,the user's interests and hobbies are subject to change at any time due to the passage of time,and the diversification of the recommendation algorithm is increased by increasing the time factor.For the collation of the mall dataset,obtain the applicable user behavior data,and compare the recommendation algorithm before and after the performance index such as accuracy,recall rate,coverage rate,diversity,etc.,and prove the recommendation algorithm in various aspects.The correctness of diversity is improved.