Against the Trend-An tentative Data Analysis Method using Classical Regression against Machine Learning Approach

Abstract

The machine learning approach is a new hot topic in recent years that are widely used in different sections, including industries, economy, disaster prediction and politics. After decades’ of development, the available machine learning algorithms are numerous and diverse. Traditional methods such as regression, classical statistical methods, are unfortunately laid aside as non-mainstream. This paper tries to compare the classical regression with machine learning algorithm as classifier. Typical machine learning algorithm support vector machine (SVM) is compared with the classical regression. The classical regression is modified to tailor as classifier. Confidence interval and credibility of prediction from regression is developed to evaluate the prediction uncertainty. Benchmark data from public database is used to demonstrate the performance. The results showed that regression exhibits an efficient computational cost with comparative accuracy.