earticle

영어

With the change in population structure, small-scale households have rapidly progressed, and since 2015, single-person households have emerged as the main household type. Along with the change in the structure of the household type, health management of single-person households is on the rise. Therefore, this study analyzed the factors influencing the classification of unmet medical care by household type using the 13th data of the Korea Welfare Panel Study (KOWEPS) in 2018. Considering that the proportion of unsatisfied medical experience is unbalanced data, which is 0.8% of the total sample, the performance of the classification analysis algorithms was compared after correcting the sample using the resampling method. As a result of the study, it can be said that the use of the random oversampling technique is superior to that of the random undersampling. In addition, in the recent classification analysis, the performance of classification analysis has been improved by using an ensemble technique such as XGBoost. In this study, it was difficult to find a difference from the classification analysis performance of the traditional classification analysis method such as logistic regression or SVM. As in the sample in this study, in a situation where 99% of the unsatisfied medical care workers are unexperienced, even if the prediction results using the algorithm are all unexperienced, it has a loophole that the accuracy reaches 99%. Therefore, in this study, in order to improve the accuracy of classification of unmet medical experience, a more valid model was derived by solving the imbalance problem and performing classification analysis. This study is meaningful in that it proved that the analysis using the original data has limitations in samples containing multidimensional unbalanced data, such as unmet medical experience.