Meet the Winners of UmojaHack Africa 2022: Faulty Air Quality Sensor Challenge (Beginner)
Meet the winners · 7 Apr 2022, 15:31

The UmojaHack Africa 2022 Beginner Challenge winners are here! We caught up with Maryam Afolabi, Vincent Njonge and Khaireddine Medhioub on their different strategies and winning solutions in the Faulty Air Quality Sensor Challenge (Beginner Challenge), hosted by Zindi as part of the recent UmojaHack Africa 2022 hackathon.

The objective of this challenge was to develop a binary classification model to predict whether an air quality sensor is faulty or not for AirQo, an air quality monitoring and research organisation in Uganda. The solutions will help AirQo to keep its network of air quality sensors in Kampala, Uganda operating at its best, and help improve health outcomes in the city and beyond.

1st Place: Maryam Afolabi (Mdda) from Nigeria

Please introduce yourself?

My name is Maryam Afolabi. I’m a student at the University of Abuja pursuing a Bachelor’s Degree in Veterinary Medicine. I love programming and practising machine learning.

How did you prepare for UHA22?

I took time to practice coding and learning other relevant subjects that would equip me for the hackathon. I also set up a good working environment for the hackathon.

Please explain your solution and what set your winning solution apart from others.

What I believe made me stand out was my feature engineering and model building. I had to drop the relative_humidity and temperature columns from both the train and test datasets since they missed a lot of values. I also extracted the date from the date column by extracting the month, year, hour, and other relative values. After that, I applied feature encoding on the date column.

For the model building, I employed the Bagging Classifier and the Light-GBM Classifier. I used the Light-GBM as the base estimator for the Bagging Classifier.

What do you like about Zindi?

I love the fact that Zindi is a platform where everyone is allowed to showcase themselves. It’s a place where people are allowed to be themselves.

Do you have any words of encouragement for others, or advice that has helped you?

Believing in myself and having an excellent mentor are two things that have helped me get this far - I would advise beginners to start taking part in competitions and get mentors that will push them forward.

2nd Place: Vincent Njonge (Pynux) from Kenya

Please introduce yourself?

My name is Vincent Njonge. I am a Data Analyst at Macro-Eyes. I have a Bachelor’s Degree in Computer Science and Mathematics from Jomo Kenyatta University of Agriculture and Technology.

How did you prepare for UHA22?

I have been taking challenges on the Zindi platform for a while now which honed my skills.

Please explain your solution and what set your winning solution apart from others.

I took time in understanding the problem and the dataset. I had to do thorough feature engineering - since there were many missing values, I decided to drop them.

What do you like about Zindi?

Zindi gives us a healthy platform to gain skills and connect to other Data Scientists.

Do you have any words of encouragement for others, or advice that has helped you?

Aspiring data scientists should take on challenges as they come and reach out to other Data Scientists - also be adept at the ever-changing technology by reading research papers. They should also ensure that they have a strong foundation in basic programming.

3rd Place: Khaireddine Medhioub (kh01) from Tunisia

Please introduce yourself?

My name is Khaireddine Medhioub. I am a Multidisciplinary Engineering Student at Tunisia Polytechnic School. I have a diverse scientific background.

How did you prepare for UHA22?

I set up a conducive environment, and as soon as the competition started I took time in understanding the data and the problem that we are to solve. Understanding the data gave me an idea of how to go about the different features in the dataset.

Please explain your solution and what set your winning solution apart from others.

I used feature importance to list features that were most important, which led me to decide to drop the relative_humidity and temperature columns. I separated the date columns into distinct columns. I used unsupervised learning (K-means) to separate classification.

What do you like about Zindi?

It’s a platform that gives the opportunity to work on different projects with different scopes and subjects to address - it has helped me improve and develop my skills within a short time.

Do you have any words of encouragement for others, or advice that has helped you?

For the beginners who are just starting out, the important thing you should embrace is learning, not winning. By practising, you will gain more confidence and get better results.