Article Preview
TopIntroduction
The Fourth Industrial Revolution, which refers to digital transformation, created a digital economic system in which economic activities were carried out as a major factor in digital input such as digital technology, infrastructure, and data. As the COVID-19 pandemic promoted a non-face-to-face society centered on digital technology, the digital era came as digital technology was brought into a mainstream way of life, not an instrumental dimension for efficiency. Changes in the current economic environment require the use of strategic digital technology by companies (Jeong, 2019).
Through this, the corporate mindset is changing from product-oriented thinking to solving problems and inconveniences experienced by customers. Key accelerating factors include Connectivity and data collection (Lee et al, 2021). D.N.A (Data, Network, AI) is a key technology in the digital era and is rapidly growing around new businesses of companies using D.N.A. The existing resource and capital-oriented business model is not only changing to a data and artificial intelligence based business model, but also developing differentiated services using machine learning in existing industries such as finance and energy. In addition, D.N.A. is used in corporate activities in various ways such as reducing repetitive tasks or reducing existing costs by using it as a reference for important decisions(Jeong & Kim, 2021).
However, the Korea Institute for Information and Communication Policy suffers from a lack of quality data, difficulty in specifying tasks, understanding AI technology, and lack of internal talent in preparation, model development, and service operation after introduction. In a survey of 152 AI demand and supply companies in Korea, 34% of respondents chose a shortage of internal manpower as an obstacle to the introduction of AI technology (total %: 200%)(Lee&Kim, 2021). Also, Glue Coding, a simple task of putting multiple cords together, accounts for 90% of the total work in the stage of model development using machine learning. The version management of machine learning, the heterogeneity of the development environment, and the actual environment cause inefficiency and causes machine learning project failure(Valohai et al, 2021).
Automated Machine Learning means automating the steps from data preprocessing to model learning to build a model using machine learning. In automatic data processing, feature selection, machine learning algorithm selection, and parameter optimization, unnecessary repetitive tasks in place of human settings and coding can be reduced to enable efficient machine learning utilization projects(Simkek, 2019; Dawid, 2021).
Therefore, the introduction of Automated Machine Learning can contribute to the development of new business models through the universal data utilization of companies, as companies can reduce the probability of failure by reducing inefficiency in machine learning projects.
Thus, this study used python to evaluate the performance of automated machine learning that is being introduced. Random Forest and Extreme Gradient Boosting models were used to predict prediction accuracy. For comparison data, 'the German Credit Prediction' dataset(UCI machine learning, 2016), which is released on the big data platform Kaggle, was used.