AI 딥러닝을 이용한 구조방정식모형의 설명력 향상에 관한 탐색적 연구 = A Study on the Explanatory Power of Structural Equation Models using AI Deep Learning

다국어 초록 (Multilingual Abstract)

The SEM-ANN two-step approach, which combines artificial intelligence deep learning technology (ANN) with structural equation modeling (SEM), was applied to management research. However, previous studies have limitations in that it is difficult to compare the research results of SEM and ANN due to differences in performance evaluation methods and cannot present performance changes according to sample size. Therefore, in this study, we compared research results using the same performance evaluation method for SEM and ANN analysis, and collected and analyzed 1,632 samples, which are relatively large compared to previous studies. The research results are as follows. First, R² was improved with the SEM-ANN two-step approach, but the improvement effect was small. Second, the generalization performance of the research model was investigated using R² obtained by applying the trained ANN model to new data. Third, sub-datasets were created by sampling 136, 307, and 660 of the total data of 1632, respectively, and the change in R² according to sample size was compared. As a result, it was confirmed that R² improved in all sub-datasets. Fourth, it was confirmed that the larger the sample size, the better the R² of the ANN model.