InstaDeep’s Mohamed Jedidi wins prediction competition and Uber Movement Prize

AI Engineer at InstaDeep, Mohamed Jedidi, wins first prize and $7,000 in Zindi’s competition “Traffic Jam: Predicting People's Movement into Nairobi”, aimed to solve the city’s traffic problems by predicting bus ticket sales.



The competition, sponsored by Uber and Mobiticket, tasked the contestants with creating a predictive model that could forecast the number of bus ticket sold for buses into Nairobi from cities in up-country Kenya. As one of the most congested cities in Africa, it was not an easy task, but Mohamad most definitely stepped up to the challenge. When the official results were announced, it was revealed that the talented 25-year old had not only won, but also been awarded the ‘Uber Movement Prize’ – an honorary award given to the highest-ranking model that used data extracted from the Uber Movement platform in the most ‘meaningful’ way.

“It’s a great feeling to win and I’m naturally honoured to also win the Uber Movement Prize. I felt comfortable I knew what I was doing, but the competition was tough with a very high competence level,” says Mohamed.

He believes the competition confronted an important and true challenge Africa faces today.

“Zindi is trying to conquer our challenges and solve our problems – Africa’s problems. The problem at hand touches the core interest of every African citizen believing in the potential of Africa, and most certainly the potential of AI”, said Mohamed.

His winning model considered essential features such as weekday, weekend, public holidays, times a day, rush hour, longitudes and latitudes between cities, traffic speed, how many bus rides a day, and other variables to outline the prediction. The approach then used Light GBM and XGBoost to train the data, tuning the parameters using a grid search. In the process, Mohamed found that, for the final results, XGBoost proved to provide better scores than LGBM.

The first prize resulted in a generous monetary return of $7,000, with an additional $2,000 for winning the Uber Movement Prize. Mohamed, which has been with InstaDeep for a year and specialises in designing and implementing intelligent systems based on machine learning, deep learning, classification, cluster analysis, uncertainty quantification and visualization, already know how he wants to invest the money. “I will use the winnings to invest in my future, more specifically in a master’s degree. I want to keep developing and grow professionally, so the money is a great contribution towards further studies when I decide I want to embark on that journey again in the future”.

The competition, which saw 571 data scientists enter, lasted for four months from 6th September 2018 to 13th January 2019. The impressive win was announced on 28th January, after a thorough review of the codes and approached behind the 1st, 2nd, and 3rd place. All contestants used Uber Movement’s traffic data combined with historic bus ticket sales data from Mobiticket as the data set. The top three winners share a prize pool of $12,000 and have handed over their models for potential future use by the Uber Movement.