应用EMPCA算法对TE过程中随机及连续缺失数据进行补值的研究

IMPUTATION OF RANDOM AND CONTINUOUS MISSING VALUE IN TE PROCESS BY EMPCA

基于传统的主元分析法,利用最大期望主元分析法EMPCA(Expectation-Maximization Principal Component Analysis)对TE(Tennessee Eastman)过程的随机缺失和连续缺失数据进行补值,并使用样本矩阵的平均相对误差及负载矩阵的误差平方和作为指标进行评价。结果表明,使用EMPCA算法补值能有效对随机缺失和连续缺失数据进行补值,效果明显优于均值补值,补值计算得到的负载矩阵误差也相对较小。

Based on traditional principal component analysis, the EMPCA （ expectation maximisation principal component analysis） algorithm is applied to the imputation of random missing value and continuous missing value in Tennessee Eastman （TE） process. The average relative error of the sample matrix and the sum of squared errors of load matrix are used as the indexes for evaluation. Results show that the use of EMPCA algorithm is able to effectively conduct imputation on random missing value and continuous missing value with significantly better effect than the mean value imputation. The error of the load matrix calculated from imputation is also relatively smaller.