Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

How to become a Spatial Data Scientist?

267 views

Published on

In this webinar you learn some of the techniques and resources that can help you bolster your Spatial Data Science skills. You can watch the recorded webinar at: https://go.carto.com/webinars/spatial-expert-recorded

Published in: Software
  • Be the first to comment

How to become a Spatial Data Scientist?

  1. 1. How to Become a Spatial Data Scientist in 2020 Follow @CARTO on Twitter
  2. 2. CARTO — Unlock the power of spatial analysis Introductions Giulia Carella Data Scientist at CARTO Steve Isaac Content Marketing Manager at CARTO
  3. 3. CARTO — Unlock the power of spatial analysis Only 1 in 3 Data Scientists have significant expertise in spatial techniques
  4. 4. CARTO — Unlock the power of spatial analysis What percentage of your Data Science team has significant experience in Spatial?
  5. 5. CARTO — Unlock the power of spatial analysis https://go.carto.com/ebooks/spatial-data-science Ready to become a Spatial Expert?
  6. 6. CARTO — Unlock the power of spatial analysis Ebook Overview
  7. 7. CARTO — Unlock the power of spatial analysis https://github.com/CartoDB/data-science-book Notebooks
  8. 8. CARTO — Unlock the power of spatial analysis Power Data Science models with location data and spatial analysis
  9. 9. CARTO — Unlock the power of spatial analysis Chapter 1 What is Spatial Data Science and Why is it Important? “Spatial data science treats location, distance, and spatial interaction as core aspects of the data” (Luc Anselin)
  10. 10. CARTO — Unlock the power of spatial analysis Spatial data comes in all forms and shapes Chapter 1
  11. 11. CARTO — Unlock the power of spatial analysis Spatial dependence “Everything is related to everything else, but near things are more related than distant things.” (Tobler, 1970) ● CONTINUOUS PROCESSES Gaussian Processes (GP), covariance functions, and variograms Chapter 1
  12. 12. CARTO — Unlock the power of spatial analysis ● DISCRETE PROCESSES neighborhood structures and autocorrelation statistics (e.g. Moran’s I) Spatial dependence “Everything is related to everything else, but near things are more related than distant things.” (Tobler, 1970) Chapter 1
  13. 13. CARTO — Unlock the power of spatial analysis ● POINT PATTERNS Complete Spatial Randomness (e.g. summary statistics, G-function) Spatial dependence “Everything is related to everything else, but near things are more related than distant things.” (Tobler, 1970) Chapter 1
  14. 14. CARTO — Unlock the power of spatial analysis Chapter 2 Spatial Modeling Leveraging Location in Prediction
  15. 15. CARTO — Unlock the power of spatial analysis Spatial modelling The mean structure e.g. a function of some covariates The residual (or what is not explained by the mean structure) What we are trying to model (or the response variable) Chapter 2
  16. 16. CARTO — Unlock the power of spatial analysis ● Continuous Spatial Error Models (GP) ● Discrete Spatial Error Models (Gaussian Markov Random Fields, GMRF) ● Spatially Varying Coefficient Models ● Spatio-temporal models ● Spatial confounding ● Validation tools Spatial modelling Chapter 2
  17. 17. CARTO — Unlock the power of spatial analysis Continuous Spatial Error Models Discrete Spatial Error Models Spatial modelling Chapter 2
  18. 18. CARTO — Unlock the power of spatial analysis Continuous Spatial Error Models Discrete Spatial Error Models Spatial modelling Chapter 2
  19. 19. CARTO — Unlock the power of spatial analysis Chapter 3 Spatial Clustering and Regionalization
  20. 20. CARTO — Unlock the power of spatial analysis Uses data attributes to create classes that, via those attributes, are different while staying alike within that category ● Longitude and latitude can be included as one of these attributes ● e.g. K-means Clustering Spatial Clustering Groups together points that are close to each other based on a distance measurement ● e.g. DBSCAN, GENERALIZED DBSCAN Clustering VS Spatial Clustering Chapter 3
  21. 21. CARTO — Unlock the power of spatial analysis Clustering Regionalization Using SKATERUsing DBSCAN Clustering VS Regionalization Chapter 3
  22. 22. CARTO — Unlock the power of spatial analysis Chapter 4 Logistics Optimization with Spatial Analysis
  23. 23. CARTO — Unlock the power of spatial analysis A typical optimization model consists of the following components: ● Decision Variables e.g. whether to open a distribution center (DC) at a specific location, whether a zip code is served by a DC, or which truck will serve one customer and when) ● Objective Function e.g. costs, service level, etc. ● Constraints e.g. physical constraints (a truck cannot transport more than its capacity), business constraints (every client should not be further than 20 miles away from the closest DC) Optimization Chapter 4
  24. 24. CARTO — Unlock the power of spatial analysis Exact VS approximate algorithms Find the actual optimal solution ● e.g. Simplex Algorithm ● Google OR-Tools Exact Approximate Close as possible to the optimum value in a reasonable amount of time ● e.g. Simulated Annealing, Tabu Search ● Google OR-Tools, Python packages (e.g. simanneal) Chapter 4
  25. 25. CARTO — Unlock the power of spatial analysis Solving the Traveling Salesman Problem Given a list of cities and the distances between each pair of cities, what is the shortest possible route that visits each city and returns to the origin city? Christofides Algorithm Ant Optimization Chapter 4
  26. 26. CARTO — Unlock the power of spatial analysis It’s time for a real world example!
  27. 27. Thanks for listening! Any questions? Request a demo at CARTO.COM Steve Isaac Content Marketing Manager // sisaac@carto.com Giulia Carella Data Scientist // giulia@carto.com

×