基于改进粒子群优化神经网络的房地产市场预测

Forecasting of real estate market based on particle swarm optimized neural network

针对粒子群优化算法精度不高、容易陷入局部最优、难以满足房地产市场形势需求的问题,提出一种改进粒子群优化神经网络,并应用于房地产市场预测中,该算法将混沌引入粒子群优化神经网络算法权重和阈值的初始化与更新的过程,提高了初始样本的质量,减轻了局部极值现象,提高了算法的全局搜索能力,同时设置了躲避因子,使粒子一定程度上离开偏离真实值的区域。研究结果表明,提出的改进算法可以提高粒子群优化神经网络权重和阈值的准确性。

Particle swarm optimization （ PSO） had the defects of low precision, and that were easily to be trapped in lo-cal optimization.To solve these problems, an neural network based on improved PSO was proposed for forecasting the real estate market.This algorithm introduced chaos sequence to update the weight and threshold, which could improve the quality of samples, reduce the local optimization and enhance the global searching ability.In addition, the avoid factor was set, which could make the particles be away from low likelihood area.Simulation results showed that this al-gorithm improved the accuracy of the weight and threshold.