Research On Neural Network Classifier Based On Improved Particle Swarm Optimization

With the continuous development and updating of artificial intelligence,research on neural networks has always been a topic of great concern.BP neural network is a branch of artificial neural networks.Because it has good prediction ability,the application is also more extensive.However,the BP neural network also has its own defects and deficiencies in practical applications.The convergence speed of BP algorithm is slow,and it is easy to fall into the local minimum.To solve the problems existing in BP neural network,in order to further improve the performance of neural network classifier,it is of great significance to design a new neural network classifier.To solve the above problems,after deeply researching the existing neural networks,this paper focuses on how to overcome the shortcomings of BP neural networks and proposes an improved particle swarm algorithm SCPSO,which is compared with the particle swarm algorithm and validates SCPSO algorithm has a strong ability for searching,and uses the SCPSO algorithm to optimize the connection weights of the neural network.It solves the problem of the slow convergence speed and easy to fall into the local minimum value of the BP neural network.Based on this,it adopts the method of combining SCPSO algorithm with neural network has redesigned the neural network classifier.Through experimental comparison and analysis,it verifies that the proposed classifier has better classification performance and aims at the existing malicious web page detection problems.The classifier implements detection of malicious web pages and verifies the efficiency and accuracy of the neural network classifier based on the SCPSO algorithm.Experiments show that the BP neural network classifier optimized by SCPSO algorithm has a high value.Compared with the BP neural network,the neural network classifier proposed in this paper has significant improvement in convergence speed,stability and other aspects.It also verifies the practicability and feasibility of the classifier by analyzing the detection results of malicious web pages.