Applying machine learning to analyze the key features of transit-oriented gentrification - A case study of Taipei metropolitan area

Abstract

This study investigates transit-oriented gentrification in the Taipei Metropolitan Area by applying a novel PCA-machine learning integrated approach. Departing from traditional indicator-based methods, we leverage Principal Component Analysis (PCA) to extract key features of gentrification from socio-economic data around existing Taipei Metro stations. Spatial autocorrelation analysis (Moran's I and LISA) identifies gentrification hotspots, providing training data for supervised machine learning models (Decision Tree, Random Forest, Gradient Boosting, and XGBoost). Our analysis reveals significant current gentrification potential in districts like Zhongzheng, Wenshan, Xinyi, and Neihu, driven by socio-economic factors. Furthermore, predictive modeling of planned MRT lines indicates that areas such as Neihu and Xizhi are likely to experience increased gentrification due to enhanced accessibility. While acknowledging limitations such as data scale variations, this research demonstrates the utility of machine learning in providing spatially explicit predictions of urban development, offering valuable insights for policymakers to formulate proactive and equitable strategies for transit-oriented development in Taipei metropolitan area.