A Hybrid PCA-Stacking Framework for Multidimensional Assessment of Development Trajectories: Evidence from China's Modernization Process

Abstract

China's multidimensional modernization since 2000 necessitates comprehensive assessment frameworks. This study constructs a novel China Development Index (CDI) through integrated analytical approaches. Principal Component Analysis first distilled 14 indicators across economic, social, and governance domains into two dominant components explaining 94.39% cumulative variance: an innovation-education-governance nexus (84.88% weight) and economic fundamentals (15.12% weight). Using Nonlinear Autoregressive Neural Network (NAR) neural networks optimized via Levenberg-Marquardt algorithms (12 hidden neurons, 2-step delays), secondary indicators like service sector growth were forecasted with 0.026 MSE. Stacking fusion modeling combining Linear Regression, k-Nearest Neighbors (KNN), and Random Forest base learners then projected primary indicators, achieving 93.02 Mean Squared Error (MSE) – 7.2% lower than individual models. The entropy weight-variation coefficient method synthesized these projections into the CDI, revealing three historical phases: Founding (1949-1978), Reform (1978-2000), and Modernization (2000-2022). Longitudinal analysis demonstrates accelerated development post-2000, with projections to 2062 indicating transition toward sustainable development characterized by environmental decoupling. A Stacking classification model integrating LightGBM, XGBoost, and Support Vector Machine (SVM) achieved 95.6% accuracy in phase identification.