K-Nearest Neighbor (K-NN) based Missing Data Imputation

The performance of the classification algorithm depends on the quality of the training data. Data quality is an important factor that affects the data mining classification results. However, one problems that often found is missing data. Effect many missing data is a less optimal classification model. Because it is can deletes important information that affect the performance of the algorithm. One method used to recover missing data is to fill it, as known as imputation. This study uses the K-NN method as an imputation carried out in several cases that have different mechanisms and missing data model. On these imputed dataset then apply classification with Naive Bayes algorithm. In this study, analyzes the performance of imputation method using Naive Bayes algorithm on the basis of accuracy for handling missing data. The results, handling missing data with K-NN-based imputation can reach the accuracy of complete data in each case with a low accuracy difference.