醫療可及性是否左右房價？機器學習之迴歸樹及隨機森林模型的應用

Does Healthcare Accessibility Matter to Housing Prices? Application of Machine- Learning Decision Tree and Random Forest

台灣邁入高齡社會，面對醫療資源區域不均，住居周遭有無充足醫療資源，對長者尤其重要，可能因而左右住居選擇。本研究透過Luo and Qi (2009) 提出的進階兩階段流動搜尋法，衡量長者之三類醫療資源可及性 (醫療需求者與供給者的空間隔絕程度)：醫師、病床與救護車。再透過機器學習之決策樹以及隨機森林迴歸，在考量住宅與其他因素下，探討長者醫療可及性對大台北地區房價的影響。
隨機森林在整體表現上優於決策樹迴歸。決策樹與隨機森林皆發現：長者之三類醫療可及性因素對大台北地區房價有顯著的影響力。由於取代台北市與新北市的平均壽命長期居各縣市之首，長者主要需要的是慢性疾病的醫療資源，相較於緊急醫療所需的救護車，醫生以及病床的可及性更為重要。這些實證結果應能提供相關單位參考，讓醫療資源達到地理區位分配最適，極大化居民生活環境品質。

 

As Taiwan gradually turns into aged society, it highlights the importance of healthcare accessibility (spatial isolation between healthcare demanders and providers). This study applied enhanced two-step floating catchment method (Luo and Qi, 2009) to estimate elders’ accessibility to three types of medical resources: physicians, hospital beds and ambulances. Then machine-learning decision tree and random forest regressions are adopted to examine the impact of healthcare accessibility on housing prices in Taipei Metropolis, Taiwan.

Random forest algorithm, as an ensemble of numberous decision trees, performed better in overall model prediction. Both machine-learning algorithms revealed that three types of medical accessibility for elders have significant effects on housing prices. Given that average life expectancy tends to be much higher in Taipei Metropolis than other regions, to elders in need of medical resources for chronic diseases in daily lives, the physician and hospital bed are proven to be more important than ambulance (needed for emergency treatment). These results should provide reference for government institutions to optimally allocate medical resources and maximize elders’ living quality.