A Method of Processing Nulls and Outliers from Data of Telecommunication Clients

This article concentrates on the field of data mining and specifically on data preprocessing, formulating an innovative method of handling datasets containing nulls and outliers. A systematic approach to dataset completion involving iteration and hyperparameter settings for treating nulls and outliers is proposed, of which automatic preprocessing on original data can be realised by setting certain hyperparameters. Weighted-average calculation is adopted to null padding to substitute traditional specific-value padding, while null-padding values are creatively used for determining outlier detection ranges. Compared with conventional ones, this method is more practicable for data of telecommunication clients.