Note | Introduction | 1. Data and features | 2. Cross-validation indices | 3. Scaling within each fold | 4. The network | 5. Training and prediction helper | 6. Running the four schemes | 7. Reading a training curve | MLP takeaways | 8. From independent rows to a graph | 9. Building the station graph | 10. From rows to a station-day panel | 11. A GraphSAGE-style model | 12. Training across folds | 13. MLP vs. GCN, side by side | Overall takeaways

Introduction | 1. The data: gstat's rural-background PM10 stations | Station coordinates | 2. Preparing the covariate table | 3. Spatiotemporal indices and station-coordinate blocks | 4. Every cross-validation scheme, side by side | 5. Prediction models: xgboost and BART | 6. Comparing the schemes | Takeaways

Row parallelization

Example 1: Spatial autocorrelation of economic activity in South Korea | Design idea | Data prep | Global Moran's I | Local Moran's I and LISA map | Example 2: Population change by sex and districts | Small multiples

Example 1: Population change in the Gyeongsangnam-do region | Design idea | Data prep | A choropleth map with an inset map | Example 2: Bivariate mapping of population vs. tax | A bivariate choropleth map | Example 3: Sex ratio in Seoul Metropolitan Area | Histograms | A choropleth map | Example 4: Socioeconomic profiling | Context | data prep | PCA results | Biplot

Getting Started with tidycensuskr | 1. Understanding Korean Geographic Hierarchies | Comparison of Administrative Divisions | 2. Available census data | Data types | Query data using anycensus() | Built-in dataset censuskor | Quick Visualization

Notes | ADM2: District-level boundaries | ADM3: Neighborhood-level boundaries

Load the data | Convert polygons to point locations | Prepare a reusable modelling split | Define model specifications for every supported engine | Fit every engine for NO | Fit every engine for NO2 | Run both responses in one loop | Notes

Objectives | Navigating KOSIS Interface | Setting Download Options | Notes on Size Restriction | Downloading Data | Post-processing Downloaded Data | Assigning Proper adm2_code | Example Code for Post-processing

Example 1: Spatial autocorrelation of economic activity in South Korea | Design idea | Data prep | Global Moran's I | Local Moran's I and LISA map

Example 1: Population change in the Gyeongsangnam-do region | Design idea | Data prep | A choropleth map with an inset map | Example 2: Bivariate mapping of population vs. tax | A bivariate choropleth map | Example 3: Sex ratio in Seoul Metropolitan Area | Histograms | A choropleth map

Introduction | The Challenge | Raw Data Sources | API call parameters | Data variables | API Key | Example URL | Data Retrieval | Setting up API Access | Downloading Raw Data | Data Cleaning Workflow | 1. Administrative Code Mapping | 2. Tax Data Processing | 3. Population Data Processing | Transforming to Tidy Format | 1. Convert Each Dataset to Long Format | 2. Combine into Single Tidy Dataset | The Result: A Tidy Dataset | Example Usage | Notes on adm2_code changes

Computational grids | Types of computational grids and their generation | par_pad_grid(): standard interface | par_pad_balanced(): focusing on getting the balanced clusters | Random points in NC | Visualize computational grids | Generate regular grid computational regions | Split the points by two 1D quantiles | Merge the grids based on the number of points | Different values in merge_max | par_make_balanced() | Common grid systems

Introduction | chopin workflow | Example data | Generating random points in North Carolina | Creating grids | Extracting values from raster | Hierarchical processing | Multiraster processing | User-defined functions | Caveats | Why parallelization is slower than the ordinary function run? | Notes on data restrictions

Introduction | Prepare target datasets | Hardware specification | Download and preprocess | Download | Read | TerraClimate | Preprocessing | PRISM dataset | Grid parallelization | Scaled up examples | Larger buffer sizes | Larger number of features | Finely resolved vector | Extract | Finely resolved raster | See also

Assumptions | Basic workflow | Minimize errors | Raster workflow: stacked vs file-based parallelization | Raster-Vector overlay | Customization | More tips to save time and memory | Save computing costs

Objective | Installation | Example | Random points in NC | Grid partition of NC | Targets workflow