The Research Of Pso Algorithm On Weather Information

With the popularization of information technology, all kinds of data has been filled with every corner of the Internet, now we need a method to deal with these data, it can be used for modeling the existing data and predicting the data of future, so that the previous data become more meaningful. Now there are two kinds of methods which are frequently used for processing and forecasting data, one method is neural network. Neural network is a more classic usage, but it is BP network, based on back propagation learning algorithm, which has been widely applied in neural network. BP network is an application of multilayer feed forward network based on gradient descent method, it can learn and store a lot of input-output model mapping, without having to reveal in advance the mathematical description of the mapping equation, and the topology structure of BP neural network mainly includes the input layer, hidden layer and output layer. Another method, support vector machine regression, is a kind of prediction method which appeared in recent years. Support vector machine regression can well solve the problems of the corresponding field for the good generalization and nonlinear characteristics, so now it is developing faster. The latter part of this paper will compare the accuracy of the two methods by forecasting the weather data to study the relationship between meteorological data and the disease.There exists some drawbacks of a standard BP algorithm, such as easy to form a local minimum value result in missing the global optimal, the learning efficiency is low for more times of training and slow convergence speed. So it is necessary to introduce some progressive strategies to improve its forecast results. This paper combines optimization algorithm and BP algorithm to optimize BP algorithm so as to improve the performance of BP algorithm and better to prepare for our forecast.Particle Swarm Optimization is a kind of evolutionary computation technology putting forward by Kennedy and Eberhart in1995. This paper will put forward a particle swarm optimization algorithm with dynamically changing inertia weight, the improved algorithm has some merits, for instance, it is easy to fall into local minimum and improve the prediction results.There are three progresses in our forecast model, in the first process, considering more dimensions of the data and a large amount of data, we first process the data by dimension reduce, then we use the clustering algorithm to cluster the data, after clustering, we train each kind of the data to generate the BP neural network based on PSO algorithm, to improve the efficiency of the algorithm and reduce the training time. In the first process of the prediction model, we process the clustering operation by K-means. K-means algorithm is a more classic algorithm toward clustering. we would get better clustering results by setting limit conditions to the center of the selected value.While we realize the prediction model through a combination of the improved particle swarm optimization algorithm and BP algorithm in the second part in order to obtain better prediction results. In the last part, we predict the main meteorological factors which are temperature, average humidity and barometric pressure station of the Anhui region with the thought of the above model. Through the experiment, we find that the framework proposed by us, which is based on improved particle swarm optimization algorithm, improves the prediction accuracy by comparing with other common frameworks.