舆情趋势预测中神经网络的优化算法

Neural Network Optimization Algorithm in Public Opinion Trends Prediction

介绍了基于BP神经网络的舆情趋势预测方法,比较分析了对BP神经网络模型进行优化的两种措施:遗传算法和模拟退火算法。针对这两种算法的不足,提出以遗传算法和模拟退火算法相结合的优化思路,既可以解决BP神经网络对初始样本值的依赖,同时也防止其陷入局部最小值,是对BP神经网络的双重优化,使得应用BP神经网络进行舆情趋势预测的准确性显著提升。

Describes the method of public opinion trends prediction based on back -propagation neural networks （BP NN） , and compares two measures which are used to optimize shortcomings of the BP NN ： genetic algorithms and simulated annealing algorithm. To supplement drawbacks of these algorithms, combine these two algorithms to optimize the BP NN. It can not only solve the dependence on the initial sam- pie values of the BP NN, but also prevent its falling into local minimum. It is dual optimization on the BP NN that will enhance the accuracy of public opinion trends prediction significantly.