基于改进遗传与神经网络的同事合乘匹配混合算法研究

HYBRID ALGORITHM OF IMPROVED GENETIC AND NEURAL NETWORKFOR COLLEAGUE CARPOOLING

针对现阶段单位同事合乘匹配搜索算法存在整体出行效率较低,算法发展速度慢等问题,提出一种遗传算法和人工神经网相融合的合乘匹配算法。该算法的遗传部分采用基因组的编码形式构造染色体,并在种群初始化时引入先验条件,提高算法效率;根据基因组形式设置特殊交叉点,保证染色体交叉过程中优秀基因组不被破坏;在构建人工神经网络过程中,以上述遗传算法迭代的过程代替网络训练过程,方便随时快速地调用。测试结果表明,混合的算法能够快速、高效地规划合乘组合,使合乘出行具有较高的效率。

Aimed at the problem of overall travel inefficiency and slow progress of the existing search algorithm for unit colleagues carpooling,this paper proposes a carpooling search algorithm combining genetic algorithm and artificial neural network.The genetic part of the algorithm used the coding form of the genome to construct chromosomes,and introduced a priori conditions during population initialization to improve the efficiency of the algorithm.The algorithm set special intersections according to the form of the genome to ensure that the excellent genome could not be destroyed during the chromosome crossing process.The iterative process of it replaced the network training process in the process of constructing artificial neural network,and thus could be convenient to call at any time.The test results show that the hybrid algorithm can plan the carpooling combination efficiently and make the carpooling more efficient.