Study Of The Swarm Intelligence Algorithm Based On Optimizing The Parameters Of SVM

Swarm intelligence algorithm is a method enlightened from the behaviors of biology and combined with computer technology to solve complicated problems appearing in the real life. It is easy for us to comprehend and achieve so that it has been used in many fields. The study in this paper aiming at the sand-dust storm forecasting enlarges the application fields of swarm intelligence algorithm.The high frequency of the sand-dust storm destroyed people's daily life and social production, the sand-dust storm forecasting is attracting for researchers. Support vector machine has been used to forecast the sand-dust storm, but the SVM's parameters are still selected using manual method , depending on experience , this effects the forecasting accuracy. In the background of the research mentioned above, for the first time, in this paper GA and PSO are used in the parameter's optimization of sand-dust storm forecasting model——RBF-SVM, and take GA and PSO as examples to make a deep research about swarm intelligence algorithm.The main researching contents of this dissertation are as follows:1. A deep research of GA and PSO is made and used respectively in the parameter's optimization of SVM. The results of the experiment show that the performance of PSO is better than GA.2. The constringency of PSO is analyzed, the condition of the algorithm's constringency and the relationship of the parameters is obtained. An improved PSO algorithm is proposed. aiming at the shortcoming that PSO is easy to get into part extremum, greaterωis used at the beginning , as the algorithm is runningωis decreasing according to a linear rule, this ensures the algorithm searches entirely at the beginning and searches partly in the end. The study parameters c1, c2 are selected under the constringency condition . The simulations'results show that the improved algorithm is better in the entire constringency and accuracy.Searching the best parameters of SVM using swarm intelligence algorithm comes true in this paper, this improves the accuracy of the sand-dust storm forecasting, promotes the development of swarm intelligence algorithm.