Based On Chaos Particle Swarm Optimization (pso) Algorithm Of Text Information Filtering Technology Research Network

For the past few years, the rapid development of the network technology, lead to a dramaticincrease of information content. This brings opportunity to people but also compels people spendlots of energy to look for valuable information from mass information. In addition, people alsowill be marred by rubbish information such as violence, pornography, reactionism and so on.Therefore, how to seek out the interested information and meanwhile filter out the irrelevant orillegal information more effectively and accurately, has become one of significant missions ininformation age.Information filtering technology is an effective method in solving the above-mentionedissues. Because more than90percent of network information is text information, so theresearches of this paper aim at text information filtering. The network text information filteringtechnology is discussed and studied in this paper. And the generation method of filteringtemplate based on chaos particle swarm optimization is highlighting introduced. The main workand innovation are as follows:1. Aiming at the problem of premature convergence, a modified chaos particle swarmoptimization is proposed.In the method, chaotic sequence generated by cube map is applied to initiate individualposition firstly, which strengthens the diversity of global searching. Then make the inertia weightchange with the particle fitness value to improve convergence rate. Furthermore, when thepremature convergence occurs, chaos perturbation is utilized to make the algorithm jump out ofthe local optimum. The experiment’s results indicate that the convergence rate and precision ofthe improved algorithm are obviously enhanced, and the algorithm can effectively avoidpremature convergence problem.2. Propose a filtering template generation method based on modified particle swarmoptimization.The method utilizes modified chaos particle swarm optimization to optimize feature subset,and propose an evaluation system of the particle fitness value based on similarity, classificationaccuracy and feature number. In addition, considering that the classifier need to be trained inevery generation, which would increase the algorithm complexity, so parallel computing isadopted. The results of experiment indicate that this method could extract the optimal featuresubset quickly and efficiently.3. Design and implement network text information filtering system based on abovemodified strategy.The proposed filtering template generation method based on modified chaos particle swarmoptimization is applied to the network text information filtering system. According to users’ demand the network information filtering is carried out timely and the filtering accuracy isimproved, the filtering system’s veracity and stability is guaranteed.