딥러닝 기법을 적용한 기술등급 예측 학습모델

This study aims to establish a deep learning learning model to confirm the predictability of technology grade, standardize technology evaluation, apply financial support, and present directions for education in commercial information specialized high schools. This study selected 1,035 SMEs and conducted a technical evaluation based on the technology finance guidelines established in 2021, and studied five stages: data collection, multicollinearity analysis, Pearson correlation coefficient analysis, normalization transformation, and deep learning learning model training. This study analyzed the multicollinearity, TOL, VIF, Durbin-Watson, correlation coefficient, and normalization method of technology evaluation data and technology evaluation sub-data. The performance was confirmed by training the deep learning learning model of TensorFlow Linear Regression, Ridge Regression, and XGB Regression. Among the deep learning learning model performances, it was confirmed that the performance and stabilization of the TensorFlow learning model were the best when RMSE and R2 were comprehensively considered. As a result of checking the accuracy of this study by dividing the technology grade into upper, middle, and lower, the TensorFlow learning model confirmed 89.39% accuracy when using technology evaluation data + technology evaluation sub-data. If a deep learning learning model with enhanced performance is studied in the future, it can be applied in stages to predict technology grade, but this study has limitations that are insufficient to reflect reliability and diversity by studying a small number of SMEs.