A Study on Prediction Performance Measurement of Automated Machine Learning

Abstract

In digital economics, where value creation using big data becomes important, the ability to analyze data using machine learning and deep learning technology is a key activity in corporate activities. Nevertheless, companies consider it difficult to introduce machine learning and artificial intelligence technologies because they need an understanding of the business as well as data and analysis algorithms. Accordingly, services such as automated machine learning have emerged for easy use of machine learning. In this study, the authors explored the automated machine learning service and compared the random forest and extreme gradient boosting analysis results using WiseProphet and Python. WiseProphet is used as a representative of automated machine learning solutions because it is a cloud-based service that anyone can easily access and can be used in various ways. It is contrasted with the model implemented by Python, which writes code with No coding. As a result of comparing the prediction performance, WiseProphet automatically outperformed the analysis result by parameter optimization.