Method for optimizing BP neural network based on improved bacterial foraging algorithm

Abstract

The invention discloses a method for optimizing a BP neural network based on an improved bacterial foraging algorithm. The method comprises the steps that: a neural network structure is determined, and related parameters of the neural network structure are set; the positions of bacteria are initialized; a bacterial component is converted into the weight and threshold of the neural network; the bacteria are turned over, so that a fitness value after one-time turning over is obtained, and if the fitness value becomes better, the bacteria are moved by a corresponding step length according to a turning-over direction; bacterial energy is obtained, one half of poor bacterial energy is eliminated, the other half of the poor bacterial energy is bred, and child bacteria have the same position andstep length as mother bacteria; the pheromone concentration of the current bacteria is acquired, the transition probability is calculated, and the positions of the bacteria are updated, so that an updated fitness value is obtained; and a group optimal solution is obtained, and the optimal solution is converted into the weight and threshold of the neural network. The method optimizes the weight andthreshold of the neural network, improves the performance of the neural network, and enables a prediction result to be more accurate.

Description

technical field [0001] The invention relates to the field of data mining, in particular to a method for optimizing a BP neural network based on an improved bacterial foraging algorithm. Background technique [0002] In recent years, swarm intelligence optimization algorithms have been widely used, most of which are based on higher organisms as inspiration objects. Passino proposed the Bacterial Foraging Algorithm (Bacterial Foraging Optimization, BFO) in 2002, starting from the behavior mechanism of microorganisms, simulating the perception of bacteria on the environment. Changes are achieved through competition and cooperation of bacterial populations. It provides a new idea for the study of bionic computing. As an optimization algorithm in continuous domain, BFO algorithm has the advantages of parallel search and easy jumping out of local minimum, and is widely used in image analysis and processing, controller optimization design and neural network training. [0003] In ...

AI Technical Summary

Problems solved by technology

The above method has achieved better performance than the standard bacterial foraging algorithm in terms of test functions or practical problems, but it still uses the bacterial foraging process, especially in the chemotaxis process, which first performs an arbitrary unit direction flip , and then perform multi-step forward operation, the complexity of the algorithm is not improved, and the convergence speed is not fast enough

Images