Mobile Money and Financial Inclusion in Tanzania Challenge
Cash and prizes worth $2,250 USD
Who is most likely to use mobile money? And who is most likely to use other financial services?
26 March–30 June 2019 23:59
462 data scientists enrolled, 163 on the leaderboard
published 6 May 2019, 10:36
edited 1 minute later

Did anyone start using ArcGIS ? If so, how can we use it to extract meaningful features?

It can be used for both visualisation and spatial analytics. For example, you can work out how far respondents are from each other or from their service provider. This can feed into your statistical models. Would it help if we stood up a webmap with all the available data shown for you to have a play? We can also share specific tutorials on point to point spatial anlytics etc


The Africa GeoPortal Team

I've started playing around, but still in the poking around phase.

@AfricaGeoPortal yes, a showcase of what's available would be awesome!

Additionally, I've been running into issues with ArcGIS blocking me when I try to load more than 1000 points into a layer. Can anyone confirm this gets solved when once you authenticate the GIS with valid user details?

Hmmm, documentation says the 1000 points limit is because I'm using pandas dataframes and not spacial dataframes...

Can confirm that authenticating doesn't fix this problem :D

As promised, we've pulled together some information to help show how ArcGIS can help in this competition. Check out the links below for more info:

We've put all the data available for the camp into one webmap so you can visualise it (you'll need to log in to Africa GeoPortal to view it)

We've also shown an example of hwo you can run spatial analysis to work out distance from a survey responder in the test file to their nearest ATM (you'll need to log in to Africa GeoPortal to view it)

To learn more about nearest analysis, check out this online course

Hope that helps everyone

Thanks so much for this input! Really helps to understand ArcGIS Online.

I've been playing around with similar things and have one question - when I try to recreate the analysis to calculate the nearest ATMs on the training set, using drive time as the distance metric and limiting to top 30 the required credits is close to 100,000. Even when limiting the analysis to using top 1, the required credits is around 3,000.

From what I can see, the only feasible option is to select line distance as the metric, which reduces the required credits to around 10.

Seeing as we receive 500 credits on sign up, even one driving / walking based analysis is not possible at this stage.

Am I missing something or is it simply an expensive operation?

Thanks again for the input and looking forward to hearing back :)

Additionally, when running the analysis using line distance I get an error message saying the number of features cannot be more than 5,000. Does this mean we cannot run an analysis on more than 5,000 samples at a time?

I downloaded the ArcGIS shapes for places/regions in Tanzania, Then mapped the shape to Longitude and Latitude given in the data i.e Training and was able the extract 19 regions. Now thinking of how I can create features based on this regions, may be nearest mobile agent, bank e.t.c