Spatial Data Science with R
7 likes12,952 views
The document is an introduction to spatial data science using R, discussing its purpose, methodology, and the integration of R with various GIS software. It outlines the skills required for spatial data scientists, highlights the advantages of using R for spatial analysis, and provides insight into various packages and techniques available for data manipulation and visualization. The presentation also emphasizes the importance of reproducibility in data science research.
Spatial Data Science with R
- 1. 1 Intro to Spatial Data Science with R AlĆ Santacruz amsantac.co JULY 2016
- 2. 2 About me ā¢ Expert in geomatics with a background in environmental sciences ā¢ R geek ā¢ PhD candidate in Geography ā¢ Interested in Spatial Data Science ā¢ Author of several R packages (available on CRAN)
- 3. 3 Purpose of this talk ā¢ Discuss what Spatial Data Science is ā¢ Give an introductory explanation about how to conduct Spatial Data Science with R
- 4. 4 What is Spatial Data Science? Spatial Data Scientist (n.): Statistician GIS/RS expertGIS developer Software engineer Spatial Data Scientist Spatial Data Science Data Science Spatial Person that is better in spatial data analysis than a GIS developer and better in software engineering than a GIS/RS expert
- 5. 5 Spatial Data Science All they are combined for data analysis in order to ā¦ Support a better decision making "The key word in data science is not data; it is science" Jeff Leek. Data Science Specialization. Coursera.
- 7. 7 Hacking skills ā¢ Programming languages: Python and R (and others) http://www.kdnuggets.com/2016/06/r-python-top-analytics-data-mining-data-science-software.html
- 8. 8 Why should we use R? ā¢ Free and open-source ā¢ A large and comprehensive set of packages (> 8600) ā¢ Data access ā¢ Data cleaning ā¢ Analysis ā¢ Visualization and report generation ā¢ Excellent development environments ā RStudio IDE ā¢ An active and friendly developers community ā¢ A huge users community: > 2 million
- 9. 9 Why R for spatial analysis ā¢ 160+ packages in CRAN Task View: Analysis of Spatial Data ā¢ Classes for spatial (and spatio-temporal) data ā¢ Spatial data import/export ā¢ Exploratory spatial data analysis ā¢ Support for vector and raster operations ā¢ Spatial statistics ā¢ Data visualization through static and dynamic (web) graphics ā¢ Integration with GIS software ā¢ Easy integration with techniques from non-spatial packages
- 10. 10 R classes for spatial data ā¢ Before 2003: ā¢ Several packages with different assumptions on how spatial data was structured ā¢ From 2003: ā¢ āspā package: extends R classes and methods for spatial data (vector and raster) ā¢ From 2010: ā¢ ārasterā package: deals with raster files stored in disk that are too large to be loaded on memory (RAM)
- 11. 11 R classes for spatial data SpatialPointsDataFrame SpatialLinesDataFrame SpatialPolygonsDataFrame SpatialPixelsDataFrame SpatialGridDataFrame sp package RasterLayer RasterStack RasterBrick raster package (recommended)
- 12. 12 The Data Science Process Modified from science2knowledge Reproducibility
- 13. 13 MODELAR los datos MODEL the data EXPLORAR los datos EXPLORE the data PREPARAR los datos PREPARE the data ā¢ Is this A or B or C? :: classification ā¢ Is this weird? :: anomaly detection ā¢ How much/how many? :: regression ā¢ How is it organized? :: clustering ā¢ How will it change? :: prediction "The key word in data science is not data; it is science" Jeff Leek. Data Science Specialization. Coursera. OBTENER los datos GET the data Domain expertise PLANTEAR la pregunta correcta PLANTEAR la pregunta correcta ASK the right question COMUNICAR los resultados COMMUNICATE the results
- 14. 14 ā¢ Import vector layers: rgdal, raster packages ā¢ Import raster layers: raster package ā¢ Get geocoded data from APIs: twitteR package, see example ā¢ Download satellite images/geographic data: raster, modis, MODISTools packages For this slide and following ones see code and examples in in this webpage MODELAR los datos MODEL the data EXPLORAR los datos EXPLORE the data PREPARAR los datos PREPARE the data GET the data PLANTEAR la pregunta correcta PLANTEAR la pregunta correcta ASK the right question COMUNICAR los resultados COMMUNICATE the results
- 15. 15 ā¢ Data cleaning, subset, etc. ā¢ Manipulate data with āverbsā from dplyr and other Hadley-verse packages ā¢ Spatial subset (sp, raster packages) ā¢ Vector operations: ā¢ Operations on the attribute table (sp package) ā¢ Overlay: union, intersection, clip, extract values from raster data using points/polygons (raster, rgeos packages) ā¢ Dissolve (sp, rgeos packages), buffer (rgeos package) ā¢ Rasterize vector data (raster package) ā¢ Raster operations: ā¢ Map algebra, spatial filters, resampling, ā¦ (raster package) ā¢ Vectorize raster data (rgdal, raster packages) For slides 14 - 18 see code and examples in this webpage MODELAR los datos MODEL the data EXPLORAR los datos EXPLORE the data PREPARE the data OBTENER los datos GET the data PLANTEAR la pregunta correcta PLANTEAR la pregunta correcta ASK the right question COMUNICAR los resultados COMMUNICATE the results
- 16. 16 ā¢ Descriptive statistics: central tendency and spread measures ā¢ Exploratory graphics (2D, 3D): scatter plot, box plot, histogram, ā¦ ā¢ Spatial autocorrelation: ā¢ Global spatial autocorrelation statistics: Moranās I, Gearyās C, Getis and Ordās G(d) (spdep package) ā¢ Local spatial autocorrelation statistics: Moranās Ii, Getis and Ordās Gi y Gi*(d) (spdep package) MODELAR los datos MODEL the data EXPLORE the data PREPARAR los datos PREPARE the data OBTENER los datos GET the data PLANTEAR la pregunta correcta PLANTEAR la pregunta correcta ASK the right question COMUNICAR los resultados COMMUNICATE the results For slides 14 - 18 see code and examples in this webpage
- 17. 17 ā¢ Regression: ā¢ Spatial autoregressive models (spdep package) ā¢ Geographically weighted regression (spgwr package) ā¢ Classification (Machine Learning): ā¢ Supervised: RandomForests, SVM, boosting, ā¦ (caret package) ā¢ Non-supervised: k-means clustering (stats package) ā¢ Spatial statistics: ā¢ Geostatistics (gstat, geoR, geospt packages and others) ā¢ Spatial point patterns (spatstat package) MODEL the data EXPLORAR los datos EXPLORE the data PREPARAR los datos PREPARE the data OBTENER los datos GET the data PLANTEAR la pregunta correcta PLANTEAR la pregunta correcta ASK the right question COMUNICAR los resultados COMMUNICATE the results For slides 14 - 18 see code and examples in this webpage
- 18. 18 ā¢ Static or interactive maps: tmap, leaflet, mapview packages ā¢ Interactive graphics, web apps and dashboards: ā¢ plotly (example), rcharts, googleVis (example) packages ā¢ shiny, see example ā¢ flexdashboard, see example MODELAR los datos MODEL the data EXPLORAR los datos EXPLORE the data PREPARAR los datos PREPARE the data OBTENER los datos GET the data PLANTEAR la pregunta correcta PLANTEAR la pregunta correcta ASK the right question COMMUNICATE the results For slides 14 - 18 see code and examples in this webpage
- 19. 19 Donāt forget: Reproducibility! ā¢ R code and output for examples shown in this webinar (slides 17-21) can be reproduced with this .Rmd document using RMarkdown ā¢ See this example about reproducible spatial analysis using interactive notebooks ā¢ Learn more about reproducible geoscientific research
- 20. 20 Integrating R with GIS software ā¢ QGIS: see example in this post ā¢ ArcGIS: arcgisbinding package, see example in this post ā¢ GRASS GIS: version 6, spgrass6 package; version 7, rgrass7 package ā¢ gvSIG: more info in this post ā¢ SAGA: RSAGA package ā¢ GME (Geospatial Modelling Environment): more info in this webpage
- 21. 21 References / Online resources ā¢ Bivand, R., Pebesma, E., GĆ³mez-Rubio, V. 2013. Applied Spatial Data Analysis with R. New York: Springer. 2nd ed. ā¢ R-SIG-Geo mailing list ā¢ CRAN Task View: Analysis of Spatial Data ā¢ Facebook groups: GIS with R, R project en EspaƱol ā¢ Google+ groups: Statistics and R, R Programming for Data Analysis ā¢ My blog: amsantac.co/blog.html
- 22. If you have any question feel free to contact me: amsantac.co/contact.html Thanks!