22 Feb 2019, 08:10

Meet the Winners of the Busara Mental Health Challenge!

Get insights from challenge winners! This is a special learning competition, as we are able to share the winning code publically. A big thank you to the winners who have posted their code on GitHub for you to learn from!

This competition was sponsored by the Busara Center for Behavioral Economics in Nairobi Kenya. The objective was to predict depression cases from routine survey data.

A model which accurately predicts which individuals are likely to be suffering from depression based on survey data could be used by mental health providers such as local clinics, NGOs or community health volunteers to reach out to those at risk.

Closing on 28 January, this challenge attracted over 170 data scientists!

We are pleased to introduce the top finishers: Steven Simba Irere of Kenya, Taiwo Ogundare of Nigeria, and Olaleye Eniola of Nigeria.

Name: Steven Simba Ireri

Solution on Github: https://github.com/StevenSimba/zindi_busara

Handle: Steven_Simba

Where are you from: Nairobi, Kenya

Tell us a bit about yourself.

I work with a micro-finance startup that gives instant loans to people who would traditionally be overlooked by banks due to lack of a credit history or savings. We use non traditional data and machine learning to build credit scores that helps us to determine how much we should lend to a customer and at what interest rate.

Tell us a bit about your solution and the approach you took. As expected from survey data, there were a lot of missing values in both the training and test data sets. My approach was to use linear interpolation to fill in the missing values. The winning solution was a blend of 3 models. It was a simple merging of predictions from a random forest model and two gradient boosting models.

What were the things that made the difference for you that you think others can learn from? Nothing in particular. However, I had entered the Devex sustainable development goals challenge on the final day of the competition and finished 4th. This gave me the confidence that it is possible to win this competition by joining early and then spending a couple of weeks gradually improving the score.

What are you looking forward to most about the Zindi community? With every new competition there are new data scientists joining Zindi. I look forward to more sharing and more learning.

Name: Taiwo Ogundare

Solution on Github: https://github.com/horlar1/Zindi-Busara-mental-health-predicition

Handle: Holar

Where are you from: Ogun State, Nigeria

Tell us a bit about yourself.

I am a ML and AI enthusiast, with B.Tech in mechanical engineering from LAUTECH,Nigeria. I primarily use R for data analysis and modelling.

Tell us a bit about your solution and the approach you took. My solution was a single xgboost model prediction using R caret package while adjusting the predicted probabilities threshold to 0.3(I.e greater than 0.3 = depressed)

What were the things that made the difference for you that you think others can learn from? Data cleaning and feature selection. Most importantly keeping it simple.

What are you looking forward to most about the Zindi community? Looking forward to seeing African talents coming together to solve challenges arising in Africa.

Name: Olaleye Eniola

Solution on Github: https://github.com/galileoSolution/3rd-place-solution-BUSARA-CHALLENGE

Handle: OLALEYE_ENIOLA_DSN

Where are you from: Nigeria

Tell us a bit about yourself.

I am OLALEYE ENIOLA a hardworking and result-oriented team player, a 300-level student of Systems Engineering.

Tell us a bit about your solution and the approach you took. I made my solution as simple as possible and applying gradient boosting after making the data sets useful to be fit into the algorithm.

What were the things that made the difference for you that you think others can learn from? I made sure I deleted irrelevant features and I tried boosting techniques.

What are you looking forward to most about the Zindi community? Collaboration