Feedback Incremental Learning Algorithm And Its Application In Network Information Filtering Research

The users can obtain abundant information through the network. At the same time, becauseof the openness of network information, the users also will be inevitably exposed to rubbishinformation such as violence, pornography, feudal superstition, racism and so on which haveobvious conscious tendency. It is a problem which needs to be solved that deriving from theneeded information of the users accurately and filtering out the rubbish information efficientlypeople are not interested in. Information filtering technology based on contents helps to retrievethe interested information while filtering out the illegal information. However, in the course ofnetwork information filtering, illegal texts’contents have strong timeliness and they will changeover time, background, places. Filtering template must be updated in time to reflect the changesin order to ensure performance and efficiency of the information filtering system. In the face ofnew challenge of real-time information filtering, capturing users’demand for the latestinformation, studying incremental learning method of network information filtering template,adjusting filtering template in time become people’s concerns and research’s focus and have theprofound social significance.The network information filtering technology is studied and discussed in this paper. To thedefect that fixed filtering template cannot dynamicly track real-time users’demands in thetechnology based on network information filtering. The paper puts forward two kinds of filteringtemplate incremental learning method based on the idea of feedback and applies them to thenetwork information filtering system. The specific work and innovation are as follows:1.Use the improved feature selection method to propose an improved filtering templateincremental learning methodThe method mainly collects positive and negative feedback set and combines GA toimprove feature selection method. After feature selection of positive and negative feedback set,then adjust features’weight of filtering template. The experimental results show that afterimprovement of filtering template, the overall performance of the system has been greatlyimproved.2. Propose a filtering template feedback incremental learning algorithm based on the simpleBayes classificationThe algorithm collects feedback training set and trains them. Then use filtering templatefeedback incremental learning algorithm based on the simple Bayes classification to revise theclassifier so that the classifier can track users’demand in time to improve filtering accuracy.After repeated tests, the stability and the overall performance of the classifier is relatively good.F1 value is above 80%, and the maximum value is 90.32%.3. Design and implement Network Information Filtering System combing with feedback incremental learning methodThe proposed feedback incremental learning algorithm is applied to the networkinformation filtering system to adjust filtering template feature weight. Acquire and update users’demand to optimize Bayes classifier. Finally , the network information filtering is carried outtimely and the filtering accuracy is improved.