全民健康保險制度主要目的在於能使每一位國民在同一制度下,秉持互助互惠的精神,分擔彼此的風險,但健保給付制度不健全,導致醫療資源過度浪費,財務失衡的問題極其嚴重,面對健保費用不斷高漲,避免不必要醫療資源浪費,有效評估醫療資源分配成為一項重要的評估指標。 新生兒出生體重及母親妊娠週數一直以來都是用來衡量新生兒健康程度的指標,然而新生兒的成長發育會受母親基因、養份、抗體及母親懷孕期間生理的影響,當母親存在已知的危險因子時,對新生兒的健康程度會有很大影響,且需耗用的醫療資源更多,如能提早發現、及早治療,將可減少不必要的醫療資源浪費。 本研究所建構八種住院日預測模型,以基因邏輯斯演算法結合倒傳遞類神經網路所建構的預測模型具有較佳的預測能力,其帄均測詴準確率為88.60%,ROC曲線下面積為0.8873,且經傅利曼統計檢定後,在ROC曲線下面積發現與其他模型存有顯著差異。本研究利用資料庫系統並設計案例式推理系統之使用者評估介面,住院天數評估系統方面,其中以基因邏輯斯演算法表現較佳,系統準確率為98.03%,帄均相似度為99.75%;住院費用方面,為交叉熵演算法表現最佳,系統準確率為71.26%,帄均相似度為99.79%。本研究結果可提供相關醫療機構做為臨床輔助評估之參考依據。
The main purpose of the National Health Insurance (NHI) system is to enable every citizen within the same system to uphold the spirit of mutual assistance and risks sharing. However, the lack of NHI comprehensiveness has resulted in excessive waste of medical resources and extremely serious financial imbalance. In the face of rising NHI costs, the effective assessment of medical resource allocation is an important evaluation indicator for reducing an unnecessary waste of medical resources. The birth weight of the newborn and the number of gestation weeks of the mother have always been used to measure the health level of newborns. However, the mother’s genes, nutrients, antibodies, and physiology have a determining impact on the development and growth of newborns. The mother’s knowledge of known risk factors greatly affects the health level of the newborn and most medical resources are consumed. Early detection and treatment will reduce an unnecessary waste of resources. Eight types of hospitalization prediction models were established in this study. Using genetic logistic regression algorithm combined with back propagation neural network, the predictive models constructed demonstrate the best predictive results,with the average test accuracy of 88.6% and the area under curve of ROC is 0.8873, The database system and the user interface of the case-based reasoning system were adopted in this study. As for the assessment system of hospitalization days, genetic logistic regression algorithm combine with the case-based reasoning system had the accuracy of 98.03%, and the similarity rate of 99.75%; As for the assessment system of hospitalization costs, cross entropy algorithm combine with the case-based reasoning system had the accuracy of 71.26%, and the similarity rate of 99.79%. The study findings shall serve as a reference for relevant medical institutions when engaging in clinical assistance assessment.