B2B marketing recommendation system based on improved genetic algorithm model

When facing abundant customer data information, every enterprise faces the problem of predicting customer consumption behavior. To address this issue, this study constructs a genetic backpropagation neural network model. The experiment compared the constructed model (M1) with the backpropagation neural network model (M2). Final test data proved that the average F-measure of the M1 was 0.952, the fitness value was 0.90, and the error value was 10−2, and it tended to converge after 40 iterations. The M2 tended to converge after 60 iterations, with a fitness value of 0.75 and an error value of 10−1. The average consumption time of the M2 in testing was 39.88 seconds, while the M1 was 18.98 seconds. Compared to these two models, the enhanced model demonstrates a notable reduction in consumption time and an accelerated training speed, while also exhibiting a more efficient use of time. This study is of great significance in developing the marketing recommendation systems.BPNN is composed of three parts: input, implicit, and output. This structural design renders it the most widely used feedforward network model. The various stages in BPNN are connected by weight thresholds that can store data, and the hidden layer can be a single-layer or multi-layer structure, which can be used to extract features of data information during training. A notable strength of BPNN lies in its capacity to calculate any nonlinear function, a feat made possible by its simple structure and algorithm, notable stability, and robustness.