GeoTuple a Framework for Web Based Geo-Analytics with R and PostGIS
Download as PPTX, PDF0 likes74 views
GeoTuple is a framework that provides exploratory spatial analysis tools by integrating R, PostGIS, and web technologies. It stores geographic data as points at multiple scales in a PostGIS database. Users can perform spatial queries and analysis via a web interface or API to reveal patterns in themes like land cover, population density, and election results across scales. The framework aggregates data values and allows comparing themes against each other or other variables like income through visualizations and statistical methods. Future goals include improving statistical modeling, importing more data sources, and developing focused applications in areas like redistricting and health studies.
![Code Javascript
R
pgSQL
//--------------------------------------------
src="scripts/opencpu-0.5.js"
var req = ocpu.rpc("getPoints", {
sw : bnds._southWest,
ne : bnds._northEast,
zoom : zoom, style: style, themes: themes
}, function(data) {
points = data[0];
draw(points, map, markerLayer);
#---------------------------------------------
library(opencpu)
library(RPostgreSQL)
getPoints <- function (sw, ne, zoom, style,
themes) {
df <- queryDbPoints(sw, ne, zoom, themes)
data <- list(df, min(df[,3]), mean(df[,3]),
max(df[,3]))
----------------------------------------------
SELECT ST_X(geog::geometry) AS x,
ST_Y(geog::geometry) AS y, %s
FROM grid
WHERE %s.geom @ ST_MakeEnvelope(. . .)
github.com/rhansson](https://image.slidesharecdn.com/cozzwy5ntjipyux2ggyd-signature-651ea1dca3295443a6523d6492cbd276766b516690b6b2a647f3525e72815d92-poli-171117185709/85/GeoTuple-a-Framework-for-Web-Based-Geo-Analytics-with-R-and-PostGIS-6-320.jpg)
![GeoTuple API requests
cURL (snippet)
$ /api_getpoint/json -H "Content-Type: application/json"
-d '{"lon":"-122.2729", "lat":"37.8021", "themes":"vote_trump"}'
[{ "dist": 49,
"vote_trump": 14
}]
$ /api_getpoints/json -H "Content-Type: application/json"
-d '{"sw_lon":"-121.552", "sw_lat":"38.543", "ne_lon":"-121.44",
"ne_lat":"38.61", "zoom":"10", "themes":["z", "vote_trump"], "key":"API_KEY"}'
[{ "x": -121.5152,
"y": 38.5698,
"z": 7
"vote_trump": 14
},
...](https://image.slidesharecdn.com/cozzwy5ntjipyux2ggyd-signature-651ea1dca3295443a6523d6492cbd276766b516690b6b2a647f3525e72815d92-poli-171117185709/85/GeoTuple-a-Framework-for-Web-Based-Geo-Analytics-with-R-and-PostGIS-14-320.jpg)
GeoTuple a Framework for Web Based Geo-Analytics with R and PostGIS
- 1. GeoTuple FOSS4G 2017, Boston a Framework for Web Based Geo- Analytics with R and PostGIS By Roland Hansson Nova Spatial @RolandHansson | geotuple.org
- 2. What is GeoTuple? tuple /ˈtjʊpəl; ˈtʌpəl/ noun 1. (computing) a row of values in a relational database
- 4. What is GeoTuple? Framework providing “exploratory analysis” tools in R ● Reveal patterns among geographic features ● Perform spatial analytics / statistics / analysis Integrates a set of OSS components Web app and API service Open Data ● Census ● Voting ● Environmental ● ...
- 5. How does GeoTuple work? OpenCPU ● RStudio Server Client ● Javascript (D3 + Leaflet) ● JSON API PostGIS database ● Multiple scales (resolutions) Cloud server (Ubuntu Linux / GCE) HTTP R (packages) Apache web server
- 6. Code Javascript R pgSQL //-------------------------------------------- src="scripts/opencpu-0.5.js" var req = ocpu.rpc("getPoints", { sw : bnds._southWest, ne : bnds._northEast, zoom : zoom, style: style, themes: themes }, function(data) { points = data[0]; draw(points, map, markerLayer); #--------------------------------------------- library(opencpu) library(RPostgreSQL) getPoints <- function (sw, ne, zoom, style, themes) { df <- queryDbPoints(sw, ne, zoom, themes) data <- list(df, min(df[,3]), mean(df[,3]), max(df[,3])) ---------------------------------------------- SELECT ST_X(geog::geometry) AS x, ST_Y(geog::geometry) AS y, %s FROM grid WHERE %s.geom @ ST_MakeEnvelope(. . .) github.com/rhansson
- 7. Point Database Grid starting at 250 meters interval Each point holds multiple values - themes
- 8. Database Themes Land Cover classification Population Density
- 9. Why points? ● Integrate data from different spatial units: ○ Census blocks, zip codes, voting districts, etc .. ● Scalable and efficient ● Aggregate at desired resolution Points (table) + R (data frame) =
- 10. That’s a lot of points! Yes, but: San Fransisco: 20 * 15 km 3,000 total points 3000
- 11. Spatial Aggregation - Multiple Scales (zoom levels) 250 meters (14) 500 m (13) 1 km (12) ... + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + Equal Area Equidistant Flexible Projections
- 12. Data Aggregation - Values Frequency (Land Cover class) A B C A A = 10 20 30 40 25 = 85 10 20 30 40 50 50 100 100 + = Weighted Average (Income per Capita) (Population Density + Income) Average (Population Density)
- 13. 250m -> 500m -> 1km Census Block -> Bl. Group -> Tract Information loss!
- 14. GeoTuple API requests cURL (snippet) $ /api_getpoint/json -H "Content-Type: application/json" -d '{"lon":"-122.2729", "lat":"37.8021", "themes":"vote_trump"}' [{ "dist": 49, "vote_trump": 14 }] $ /api_getpoints/json -H "Content-Type: application/json" -d '{"sw_lon":"-121.552", "sw_lat":"38.543", "ne_lon":"-121.44", "ne_lat":"38.61", "zoom":"10", "themes":["z", "vote_trump"], "key":"API_KEY"}' [{ "x": -121.5152, "y": 38.5698, "z": 7 "vote_trump": 14 }, ...
- 15. Theme Operation Equal Interval Classification Data Domain
- 16. Min Average Max Data Domain
- 18. Highlight: Income > 70000 AND Elevation < 100
- 19. Income Inequality < 10km Difference > $100,000
- 21. Comparing Themes vs Income Visually: a) b) c)
- 22. Strength of relationship with Income for themes: ● Race (white) ● Voting Trump/Clinton ● Distance Starbucks ● Population Density ● Environment / pollution r = Correlation coefficient
- 23. Looking Ahead @RolandHansson | geotuple.org Better Statistical Modeling ● Spatial Autocorrelation ● Transforming non-linear relationships ● Interpolation More Data ● Import custom data ● Sustainability Index (access to public transport, amenities ...) Focused Solutions ● Improved and streamlined UI ● Redistricting, health studies Big Data ● GPU + Analytics + Machine Learning
Editor's Notes
- #22: http://andrewgelman.com/2011/04/04/irritating_pseu/ Andrew Gelman, Professor, Columbia University