遗传算法和BP神经网络的电信业产出预测

电信业产出是衡量电信产业发展状况的主要指标,其影响因素具有较强的非线性和随机性,很难用传统的预测方法进行预测。而BP神经网络方法具有的较强非线性映射能力和柔性网络结构在解决这一问题时具有明显优势。本文基于遗传算法改进的BP神经网络模型,利用我国电信市场从1998至2012年15年数据作为训练数据与测试数据应用于电信业产出规模预测。研究结果表明,基于遗传算法的BP神经网络模型预测精度明显大于传统预测方法,可为电信业的健康和快速发展提供一种新的、可量化的工具和方法。

The output scale of telecom industry is the main indicator to measure the development status of the telecom industry. As the influential factors have strong nonlinearity and randomness, it is difficult to predict it by the conventional prediction method. The BP neural network method has obvious advantages in addressing this issue because of its strong nonlinear mapping ability and flexible network architecture. This paper is based on the BP neural network model improved by genetic algorithm, by the use of China＇ s telecom market data from 1998 to 2012 as training data and test data to forecast the scale of telecommuni- cations market. The results show that prediction accuracy of BP neural network based on the genetic algorithm is significantly greater than traditional forecasting methods, which can provide a new, quantifiable tool and method for the healthy and rapid development of telecom industry.