Fuzzy Genetic Algorithm And Its Application Research In Network Information Filtering System

With the increasingly large number of network information on the internet, the demand on the high quality of network information becomes higher, more and more the new technology is being in progress, because the network information has strong dynamic nature, which is mainly represented on the dynamic change of webpage contents and link address, so, about the content filtering of the network information, mostly processing the dynamic information flow on webpage. When processing the numerous of redundant information data,the demand of users to dynamic information must be met ,and filter out uninterested information of users. The network information system collects or clears away certain text message from the dynamic dataflow .The main function of network information system is to filter out the harmful information on the network, mostly reflects on pornography, violence, crime and so on.According to the network information filtering based on tradition genetic algorithm, the new genetic algorithm which is proposed in this project, this algorithm is called fuzzy genetic algorithm. Because some uncertain factors occur on the network information system, fuzzy genetic algorithm can adjust these uncertain factors properly, including parameters values and weight values and so on. Based on the traditional genetic algorithm, analyze the characteristics of text training set, improve text classification performance, add the fuzzy methods to the system adjust parameters and weights ,makes these factors varied with system environment change . The research theories of fuzzy genetic algorithm are embodied in the following four aspects:1. Fully analyze network information filtering modelNetwork information flitering system structure mostly reflects in data packet capture, feature selection, optimization algorithm, classification algorithm, each part of key technologies of this system are analyzed, expound its advantages and disadvantages, put forward betterment methods ,fully improve the effects of network information filtering.2. Document training sets processing technologyIn past, as far as document set is concerned, training texts in corpus were treated fairly, and ignored the importance of the each text, in order to improve those disadvantages, the effects of training text to classification should be emphasized, text should all be treated differently according to the importance of document. For document training sets processing are mostly reflect in divide text paragraph and set document weight. Text paragraph divide makes text divide into the first paragraph,the middle paragraph,the ending paragraph and different paragraph integrated, according to each paragraph different effect, choose out important paragraph to training, The results show that the first paragraph and ending paragraph integrated better than single paragraph. Using of paragraph training replaces of the whole text training, it is not only greatly relieves time complexity, but also relieves the system burden, and improves the system operating efficiency.In the Large-scale text corpus, each paragraph has different contain, in classification, the effects are also different, it can be called the importance of the document. In this project, one of the important research work is the importance of document, namely document weight, fuzzy computation methods are applied to this system, according to the classification document corpus and classification effects, using fuzzy computation methods adjust document weight ,makes the important documents obtain fully application ,further improve the quality of document corpus.3. Genetic algorithm parameters adjustmentThe genetic algorithm parameters contain group size , selection probability,crossover probability,mutation probability and evolution times and so on .Those parameters will greatly affect genetic algorithm quality. In past, genetic algorithm parameters are all fixed value, those parameters value can not varied with genetic environment change. Therefore, in order to improve those parameters, in the project, those parameters value can be adjusted by the fuzzy methods, it mostly reflects in crossover probability and mutation probability will do different adjustment in different genetic environment. Genetic algorithm can get well useful in the optimization stage, and obtain the best solution of genetic algorithm.4. Fuzzy adjust feature item weightAccording to feature selection method, feature item weight can be computed, and then using genetic algorithm optimizes training feature item, choose out the most meaningful feature item. Because weight result will affect the classification effect to a large extent, therefore, adjust feature weight appropriate according to the system environment. In this project, put forward a kind of improve methods of feature item weight, feature item discrimination method is studied in order to adjust feature item weight, and then rebuild classification template. The results show that the adjusted experiment effects better than before.