Abstract: In order to improve the prediction accuracy of short-term traffic flow, in view of parameters optimization problem for Least Squares Support Vector Machine（LSSVM）, this paper proposes a short-term traffic prediction model based on Artificial Fish Swarm Algorithm（AFSA） and LSSVM（AFSA-LSSVM） which LSSVM parameters are optimized by improved AFSA, and the simulation experiment is carried out based on short-term traffic flow data. The simulation results show that the proposed model can obtain better LSSVM parameters and can more accurately reflect the changes of short-term traffic flow, and improves the short-term traffic prediction accuracy compared with the reference model, and it provides a new research method for the nonlinear short-term traffic flow prediction.

Key words: short-time traffic flow, least squares support vector machine, artificial fish swarm algorithm, time series

摘要： 为了提高短时交通流量的预测精度，针对最小二乘支持向量机（LSSVM）参数优化难题，提出一种人工鱼群算法（AFSA）和LSSVM相结合的短时交流量预测模型（AFSA-LSSVM），通过采用AFSA优化LSSVM参数，并采用具体短时交通流量数据进行仿真实验。仿真结果表明，相对于参比模型，AFSA-LSSVM可以获得更优的LSSVM参数，能够更加准确地描述短时交通流量变化趋势，提高了短时交通量的预测精度，为非线性短时交通流量预测提供了一种新的研究思路。

关键词: 短时交通流量, 最小二乘支持向量机, 人工鱼群算法, 时间序列