Partial Weighted K-nearest Neighbor Classification of Incomplete Data

Data classification is a hot topic in machine learning research. Most classification algorithms need complete data sets, but missing data widely exists in our life. It is necessary to classify the data sets with missing data. Generally speaking, the existence of missing data will affect the classification effect, so the processing method of missing data largely determines the quality of the classification algorithm. In this paper, K-nearest Neighbor (KNN) algorithm is improved with the help of Patial Distance Strategy(PDS)to be applied to incomplete data sets. The processed data set is visualized. The classification effect is compared with the mean imputation and 0-value imputation, and the different features of the data set are weighted in various ways. The results show that our method has high accuracy on incomplete data sets.