DATASET MISSING VALUE HANDLING AND CLASSIFICATION USING DECISION TREE C5.0 AND K-NN IMPUTATION: STUDY CASE CAR EVALUATION DATASET

Data mining is a data analysis process using software to find certain patterns or rules from a large amount of data which is expected to find knowledge to support decisions. However, missing value presence in data mining often lead to loss of information. Information loss inside dataset such car evaluation can result in poor predictive models. The purpose of this study is to improve the performance of data classification with missing values precisely and accurately using Decision Tree C5.0 and k-NN Imputation. The test method is carried out using the Car Evaluation dataset from the UCI Machine Learning Repository. RStudio and RapidMiner tools were used for testing the algorithm. This study will result in data analysis of the tested parameters to measure the performance of the algorithm. Using test variations: 1. Performance at C5.0, C4.5, and k-NN at 0% missing rate. 2. Performance on C5.0, C4.5, and k-NN at 5-50% missing rate. 3. Performance on C5.0 + k-NNI, C4.5 + k-NNI, and k-NN + k-NNI at 5-50% missing rate. 4. Performance on C5.0 + CMI, C4.5 + CMI, and k-NN + CMI at 5-50% missing rate. The results show that C5.0 with k-NNI produce better classification accuracy than other tested imputation and classification algorithms. For example, for 35% missing in the dataset, this method obtains 93.40% in validation accuracy and 92% accuracy in the test. C5.0 with k-NNI also offers fast processing time compared with others methods.