隨著全球少子化趨勢加劇，嬰兒健康問題尤為重要，早產與低出生體重是導致新生兒死亡和發展障礙的主要因素，對家庭、社會及醫療系統造成重大負擔，且近年來，臺灣亦面臨生育率持續下降的問題，進一步加劇了新生兒健康議題的重要性，因此，如何有效預測並降低早產與低出生體重的發生，成為當前研究的一個重要方向。本研究共設計了兩階段實驗，透過分析孕產婦和新生兒的住院臨床資料，利用機器學習方法找出導致早產和低出生體重的潛在因素，實驗一針對新生兒早產的預測模型，實驗二針對新生兒低出生體重的預測模型。結果顯示，極限梯度提升樹在預測新生兒早產和低出生體重方面表現最佳，透過模型結果，最終建立新生兒健康的臨床決策系統，以作為早期識別高風險新生兒的工具，使醫療專業人員能及時進行干預和護理，提供更為精準的醫療資源配置，以減少不良健康結果的發生。;With the intensification of the global trend of declining birth rates, infant health issues have become particularly important. Premature birth and low birth weight are major factors leading to newborn mortality and developmental disabilities, imposing significant burdens on families, society, and healthcare systems. In recent years, Taiwan has also faced continuously declining birth rates, further emphasizing the importance of newborn health issues. Therefore, effectively predicting and reducing the occurrence of premature birth and low birth weight has become a key focus of current research. This study designed two-stage experiments, analyzing clinical data of pregnant women and newborns to identify potential factors leading to premature birth and low birth weight using machine learning methods. The first experiment focused on a predictive model for premature birth, while the second experiment focused on a predictive model for low birth weight. The results showed that extreme gradient boosting trees performed best in predicting premature birth and low birth weight in newborns. Based on the model results, a clinical decision support system for newborn health was established. This system serves as a tool for early identification of high-risk newborns, enabling healthcare professionals to intervene and provide care promptly, allocate medical resources more accurately, and reduce the occurrence of adverse health outcomes.
