基于优化最大偏差相似性准则的KNN缺失数据填充算法

KNN Missing Data Filling Algorithm Based on Optimized Maximum Deviation Similarity Criterion

根据高速公路交通数据的特点,采用基于最大偏差相似性准则(MDSC)与KNN填充算法对缺失交通数据进行填充。针对KNN填充算法可能产生伪邻近点问题,提出利用MDSC对不完整的交通数据中缺失的属性样本和完整值数据样本进行聚类,以避免伪邻近点发生;并利用基于骨干粒子群算法对MDSC参数优化。实验结果表明:基于优化MDSC的KNN填充算法的RMSE值更小,效果更优。

A KNN missing data filling algorithm based on improved maximum deviation similarity criterion is proposed.Considering the characteristics of expressway traffic data,the missing data matching algorithm based on the maximum deviation similarity criterion and KNN is confirmed to fill the missing traffic data.For the problem that the KNN filling algorithm will produce the nearest neighbor noise(pseudo-neighbor),it is proposed to use the maximum deviation similarity criterion to cluster the complete value data samples for the missing attributes in the incomplete traffic data to avoid the pseudo-neighbors.occur.Among them,the key parameter selection problem of MDSC algorithm is based on the backbone particle swarm optimization algorithm to optimize the MDSC parameters.The simulation results show that the missing RMSE value of the missing traffic data based on the improved MDSC KNN filling algorithm is smaller and the effect is better.