17 May 2019, 14:15

Meet the winners of the Sea Turtle Rescue: Error Detection Challenge!

Get insights from challenge winners! A special thank you to the winners for their generous feedback.

Closing on 29 April 2019, the Sea Turtle Rescue Error Detection Challenge was the fifth Zindi challenge to close. The objective of this competition was to create a machine learning model to help Kenyan non-profit organization Local Ocean Conservation identify potential errors and anomalies in their sea turtle rescue database. This challenge attracted over 200 data scientists from across the continent and around the world, of whom over 53 made submissions and entered the leaderboard.

We are happy to introduce the top two winners of the competition: János Sávoly of Hungary and Emmanuel Onwuegbusi of Nigeria!

Name: János Sávoly (1st prize)

Zindi handle: CacoS

Where are you from?

Budapest, Hungary

Tell us a bit about yourself.

I am a mathematician. I graduated from ELTE (Budapest). I am currently working at the Research Institute of Agricultural Economics (Budapest) where I create and maintain mathematical models for the impact assessment of the Common Agricultural Policy (CAP) and work on other data science related tasks. This is the place where I developed an interest in machine learning.

Tell us about the approach you took.

My approach was a fairly simple one, I treated this problem as 25 (number of columns in the data, except Rescue_id) separate binary classification tasks (error - no error). I used xgboost to solve these classification problems. The programming was done in R.

What were the things that made the difference for you that you think others can learn from?

This problem was not a standard one so the right formulation made a big difference. My solution relied heavily on hyperparameter tuning (Bayesian optimization, 5-times CV stratified by the year of the bycatch) and on hand crafted features like the number of extra spaces in TurtleCharacteristics, pairwise checks of some features (1, if they both contain NA or are both not NA, 0 otherwise, e.g. CaptureSite-Landingsite) and features based on Date_Caught like weekdays, year etc. I trained the models on the whole dataset with the tuned parameters. I would like to give a shoutout for the people on the mlr project (https://mlr.mlr-org.com/). Thank you for creating this great ecosystem of packages!

Name: Emmanuel Onwuegbusi (2nd prize)

Zindi handle: EmmaMichael

Where are you from?

Anambra, Nigeria

Tell us a bit about yourself.

I am a data scientist and I am passionate about applying artificial intelligence to solve Africa and the world’s challenges.

What were the things that made the difference for you that you think others can learn from?

Going through the 'info page' of the competition to understand what was expected before coding was key.

What are the biggest areas of opportunity you see in AI in Africa over the next few years?

Artificial intelligence in healthcare, agriculture and fintech.

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

I look forward to seeing a very vibrant community where data scientists in Africa and the world can collaborate to solve Africa’s and the world’s challenges.

What are your thoughts on our winners' feedback? Engage via the Discussions page or leave a comment on social media.

Competition sponsors

This competition was sponsored by Local Ocean Conservation (LOC), Temple Point Resort, and individual donors of LOC.

LOC is a private, not-for-profit organisation committed to the protection of Kenya’s marine environment. LOC supports the communities and coastal areas in Watamu and Diani, Kilifi County with marine conservation and community development projects.

The Temple Point Resort is a premier holiday beach resort located at the end of a headland between the Indian Ocean and Mida Creek.