Efficiency Comparison in Prediction of Normalization with Data Mining Classification

Abstract

In research project, efficiency comparison study in prediction of normalization with data mining classification.The purpose of the research was to compare three normalization methods in term of classification accuracy that the normalized data provided: Z-Score, Decimal Scaling and Statistical Column.The six known classifications: K-Nearest Neighbor, Decision Tree, Artificial Neural Network, Support Vector Machine, Naïve Bayes, and Binary Logistic Regression were used to evaluate the normalization methods.The six studied data sets were into two groups.Those data sets were data sets of White wine quality, Pima Indians diabetes, and Vertebral column of which data were 1-5 variables of the outlier coefficient of variation and data sets of Indian liver disease, Working hours, and Avocado of which data were 6-10 variables of the outlier coefficient of variation.The result of comparison White wine quality and Vertebral column, the best efficiency method had many methods in a non-systematic way.For the data set of Pima Indians diabetes and Indian liver disease, Statistical Column and classification by K-Nearest Neighbor was the best efficiency.For the data set of Working hours, Decimal Scaling and classification by K-Nearest Neighbor was the best efficiency.For the data set of Avocado, Statistical Column and classification by K-Nearest Neighbor, Z-Score and Decimal Scaling and classification by Binary Logistic Regression were the best efficiency.All of normalization and classification methods, Statistical Column and classification by K-Nearest Neighbor was the best efficiency by precision.