Worsted roving forecast based on genetic algorithm and BP neural network

Due to slow rate of convergence and falling into part minimums easily in BP algorithm,and wide search space,high search efficiency and big robustness of genetic algorithm(GA),a new improved genetic BP algorithm combining the two was put forward to model the fore-spinning process and forecast the quality. The weight and threshold matrix of the BP network were formed a string in an orderly way as the chromosome of GA.Through the operations of selection,crossover and mutation,they were optimized and used as the initial matrix of BP model to do second training.The verification for the same data indicates: the pure BP models can not achieve the expected precision or fall into part minimums,the models after optimization of GA all have fast convergence and achieve the expected precision.The relative mean error percent(MEP) between the forecast and the measured value of the 20 groups of testing samples are reduced to 2.55% and 2.23% from 3.56% and 3.48% respectively;the correlation coefficient between them are also improved.