使用泰国的地理空间数据预测贫困.docxVIP

CONTENTS TOC \o 1-2 \h \z \u TABLES AND FIGURES iv ABSTRACT v INTRODUCTION 1 LITERATURE REVIEW 3 Using Precompiled Geospatial Data for Predicting Socioeconomic Indicators 3 Poverty Mapping in Thailand 4 DATA 6 Satellite Data 6 Crowd-Sourced Geospatial Database 7 Poverty Data 7 Reference Period 8 METHODS 8 Generalized Least Squares 8 Neural Network 8 Random Forest 9 Support Vector Regression 10 ANALYTICAL RESULTS 10 Preliminary Analysis 10 Using Machine Learning Algorithms to Predict Income-Based Poverty Rate 11 Using Machine Learning Algorithms to Predict Multidimensional Poverty Index 15 CONCLUSION 18 REFERENCES 19 APPENDIX 25 TABLES AND FIGURES TABLES 1 Technical Specification of Data Obtained from Google Earth Engine 6 2 R Packages Used in Machine Learning Computations 10 3 Adjusted R2 Values for Random Forest-Based Models 18 A1 List of Variables Obtained from Geospatial Data of 2015 25 A2 List of Variables Obtained from Geospatial Data of 2017 28 FIGURES Proportion of Thailand’s Population Living Below its National Poverty Line 1 The Human Achievement Index in 2017 2 Illustration of a Sample Neural Network 9 Comparison of Average Root Mean Squared Error Obtained from Four Machine Learning Algorithms 11 Scatter Plot of Published and Predicted Income Poverty Rates, 2015 11 Scatter Plot of Published and Predicted Income Poverty Rates, 2017 12 Results of Variable Importance Computation 13 Results of Minimal Depth Computation 14 Comparison of Average Root Mean Squared Error Obtained from Four Machine Learning Algorithms, 2017 15 Scatter Plot of Published and Predicted Multidimensional Poverty Index, 2017 15 Result of Variable Importance Computation 16 Result of Minimal Depth Computation 17 ABSTRACT Poverty statistics are conventionally compiled using data from household income and expenditure survey or living standards survey. This study examines an alternative approach in estimating pov