Abnormal data processing in chemical production
Abstract:This thesis for actual chemical production process of soft measurement technology,including a large field of observation data, any abnormal data (wild) may cause the model prediction,even completely failed,so for measuring data pretreatment is very important.With delayed coking process coke yield soft measurement model,for example,consider multivariable clustering analysis for abnormal data of identification method,identification and interpretation of these abnormal data for subsequent modeling results abnormal samples.
This paper summarizes and discusses about the data mining, the outliers,clustering algorithm of existing research results.And introduces in detail the clustering algorithm based on density and neural network,and delayed coking process coke production rate of the soft measurement model,the original data component analysis and neural network, and then processed data of composition analysis and neural network analysis.Finally comparing with them.
Keywords：Clustering analysis;outliers；Abnormal data；
