基于改进ICSO-LSTM方法的城市交通预测研究

Research on Urban Traffic Prediction Based on Improved ICSO-LSTM Method

为进一步提高城市道路交通流量预测水平,降低城市拥堵度,提出一种基于改进鸡群算法优化长短期记忆神经网络的城市交通预测方法——Improved Chicken Swarm Optimization Long Short-Term Memory(ICSO-LSTM)。首先,针对鸡群算法存在收敛速度快、容易陷入局部最优的问题,从非线性递减的公鸡位置更新、加权的母鸡个体和自适应的跟随系数优化三个方面进行优化;其次,建立了ICSO优化LSTM网络参数的预测模型;最后,将ICSO-LSTM用于交通流量预测中,仿真实验表明,该方法在城市交通流量方面具有较好的预测效果。

In order to further improve traffic flow prediction on urban roads and reduce urban congestion,an urban traffic prediction method based on an Improved Chicken Swarm Optimization Long Short-Term Memory(ICSO-LSTM).Firstly,the Chicken Swarm Optimization is optimized in three aspects,namely,nonlinear decreasing rooster position update,weighted hen individual and adaptive following coefficient optimization,to address the problem that the Chicken Swarm Optimization has a fast convergence rate and easily falls into local optimum;secondly,the prediction model of ICSO optimized LSTM network parameters is established;finally,ICSO-LSTM is used in traffic flow prediction,and simulation experiments show that the method has good predictive performance in urban traffic flow prediction.