Research And Implementation Of Adaptive Network Information Filtering System Based On Annealing Genetic Algorithm

As a systematic approach, information filtering technology could match the user demand with the dynamic information flow from which we extract the personalized information of user demand and send it to the user. Current research on information filtering is about how to obtain and express the information, how to build the user template file with learning algorithm, how to calculate the information similarity and so on.Genetic Algorithm is applied to machine learning and combinatorial optimization whose advantages beyond other methods. Therefore, in this paper, on the basis of comparing with the relevant text classification algorithms, an information filtering system model based on Genetic Algorithm is put forward which is around the two main target of information filtering, the accuracy and filtering rate; Considering that the the premature disadvantage of the Genetic Algorithm, the lower match efficiency during the whole matching process and the problem of updating the user interest template and the document database during the training process, a series of improved measures are put forward based on which the network information filtering is carried out in this paper finally. Specific works of this article are as follows:1. An in-depth research of the key technologies of network information filtering and the related filtering modelThe general model of information filters and classification algorithms are dicussed at first and analysing the problems existed in the current information filtering system. Then, we focus on the network data's acquisition and representation, the calculating method of the featuer weights as well as the matching and classification algorithms and the key match and feedback techniques.2. Introduced the information filtering with genetic algorithm to generate the filtering template.After comparing and reasearching the traditional text classification techniques and analyzing the characteristics of genetic algorithm and its application, a text classification and information filtering model construction method based on improved genetic algorithm is put forward in this chapter, namely the genetic training.During the genetic training, we combine symbol code with binary code to deal with the management of vector text, that is, considering a series of improved genetic manipulation and the similarity degree between vectors as a fitness function, the genetic manipulation after algebra must formed a template for text classification and information filtering, and the packet captured by the network packet capture module is segmented into sub-words which can be compared with template on similarity to decide the category of this text and eventually achieve the purpose of the text fitering.3. Setting up network information filtering model based on the improved genetic algorithm.Based on a full analysis of the advantages of genetic algorithm, we introduced it into the network information filtering to generate the template; For the local optimal disadvantage of the genetic algorithm, we introduced the simulated annealing genetic algorithm to regulate and improve the genetic algorithm in the structure; For the shortcomings of the genetic parameters fixed and species single, we introduced species control on the basis of age, as well as the idea of changing cross-algebra mutation rate and variance rate.4. Introduced the logical paragraphs division method based on the concept.The method was set up based on the concept dictionary.From analyzing the logic concept contained in the text, the paragraphs with the same meaing will be clustered, and the logic levels which were established on this logic concept paragraph were used to be a basis for classification so as to consider the contribution degree of the different paragraphs to the text subject. At the same time, for the polysemy or synonym phenomenon in the matching process, synonym concept expansion and related terms expansion are introduced in this paper.5. Given a method of modifying template dynamicly with feedback documents.Type of template has a direct relationship with the merits of information filtering systems, and classification system may change frequently. To address the problem raised in the course of the study subjects using a modified type of feedback dynamic document template algorithms. The method of retraining is time-consuming, laborious and run counter to the original purpose of feedback. The true feedback should be adjusted on the generated training results, that is, the filter template is adjusted automatically in the filtering process. In view of the problem raised above, a method of modifying template dynamicly with feedback documents is put forward in this paper.6. Designed and implied a network information filtering system named NIFS.In accordance with the thinking of sub-blocks, hierarchical, as well as modular, achieved a network information filtering system. Three-tier implementation of the system filtering mechanism and SPI-based network packet interception technology to intercept and recombine the packets improved the filtering speed. Annealing genetic algorithm is employed to study on the training samples to generate the user templates which are adjusted and optimized by feedback learning to improve the filtering accuracy.