Research And Application Of Social Network Hot Spot Recommendation Algorithm

Social networking,as a social networking platform for human relationships,is an important way to provide information communication and sharing.Social networking is changing people's communication and learning in their own way.Because of the use of social networks in the vast network of users,it is the topic of the current research to dig out the hot topics that users pay attention to in the mass social data.The recommendation of hot topics in social network is data mining based on the massive topic data exchanged by users in social network.Finally,the hot topics after mining and analyzing will be displayed to users.Compared to the current social network and traditional network media,news text contains sparse,high-dimensional,network language is not normative,and with the massive exchange of information,leading to the traditional hot topic mining technology used directly in the social network has low efficiency and low accuracy problem.This paper mainly studies and analyzes some defects existing in the use of simple Bias classification algorithm and K-means clustering algorithm,and improves the above algorithm.This article mainly makes the following work on the above problems:(1)Analysis and research according to the characteristics of modern social networks in this paper,a detailed description of the hot topic detection from the network data exchange based on social exchange data preprocessing to obtain meaningful data,the processed data using classification algorithm to classify the data,then the data to complete the classification of hot topic detection by clustering algorithm.(2)The use of Naive Bayesian classification algorithm in the topic of social network in hot recommendation accuracy and efficiency issues,because the Naive Bayesian classification is generated using the method to determine the conditions to achieve the classification,so it has high classification accuracy and processing speed.However,the classification probability calculated by the simple Bias classification model may be very close,and can not be defined in specific categories.It is difficult to apply to the recommendation of hot topics.In this paper,we propose a Hot-spot Texture Selection Based on Navie Bayesian based on naive Bias classification,which is used as a text selection algorithm for hot topics in social networks.The algorithm first uses naive Bias classifier to calculate the probability of text belonging to all kinds of hot topics,and calculates the standard difference to decide the difference between the categories of the text,and determines whether the text is eliminated.(3)According to the traditional K-means algorithm on the random selection of the initial cluster center shortcomings,this paper proposes K-means initial clustering center selection algorithm based on is based on the sparsity of the feature data to initialize cluster,it first calculates the degree of aggregation points around the initial data.And the average degree of aggregation of the minimum distance formula and the adjacent correlation point to select the K point data intensity is relatively high as the clustering center.