Performance analysis of cart and C5.0 using sampling techniques

Data mining is the process of extracting the hidden predictive model from large databases. It has various methods and algorithms. Classification is a supervised method, which builds a model for predicting the new instances. Different algorithms like decision tree, neural networks, support vector machines, k nearest neighbour, Bayesian classification are available for the classification. Decision tree is the simple and most commonly used algorithm among the classification algorithms. It constructs a tree based model on the values of feature and generates rules for decision making. Samples are used for classification in order to train the model and predict the new instances. Unbiased samples can improve the performance of classification. This paper analyses the performance of CART and C5.0 algorithms using sampling techniques.

I. Introduction

Data is available in huge volume with the advent and internet technologies, rapid progress in communication speeds and it plays an important role in today's world. Extracting knowledge from this huge data store is a key challenge ahead. Data mining is used for discovering the knowledge from large databases [1]. Classification and Prediction are the most prominent areas of data analytics for predicting the new instances [2]. In Classification, dataset is divided into training dataset and testing dataset. And a model is constructed using training dataset, and then performs the prediction by applying the model on testing dataset. Here class labels of training dataset are known. Decision tree is a widely used classification algorithm, to build the hierarchical structure tree for taking the exact decision by applying the new instances.