数据处理中缺失数据填充方法的研究

Research on the Method of Filling Missing Data in Data Processing

针对数据处理中常见的缺失数据现象,研究了若干种数据填充方法.分别对拉格朗日插值法、回归分析法、灰色预测法和BP神经网络方法进行了缺失数据计算和算法对比研究.以武汉市2013年3-4月的PM2.5数据作为实验研究对象,用上述四种方法进行了缺失数据验证及估计.对比计算结果发现,对于此类数据,利用拉格朗日插值法进行缺失数据填充效果优于其他三种方法.

A number of researches on the missing data in data processing have been done for a long time. This paper compared four methods： Lagrange interpolation method, regression analysis method, gray prediction method and BP neural network. By using the above four methods, we validated the estimation of the missing data, which are the monitoring data of PM2.5 of Wuhan （March-April,2013）.The findings of Comparing the calculation results indicate that, for these data, the effect is better than the other three methods in filling the missing data by using the Lagrange interpolation.