HakkTaxi: ride-share demand prediction (Microsoft Azure APAC Hackathon)

Jakarta ride demand rises during commutes, Friday evenings, and rain. However, drivers reacted to surge signals that were already 5 to 10 minutes late. A map forecasting demand 15–30 minutes ahead could help drivers move sooner. Therefore, it could shorten surges and make earnings more predictable.

The 48-hour competition left no time to obtain private trip records. Therefore, I built a demand signal from proxy data, historical totals, weather, and static place density. It had to work across Jakarta’s varied geography. Office clusters in the central business district, university campuses, market districts, and transit hubs had different demand patterns. The model needed to capture these patterns without fitting too closely to training data.

Moreover, I had to demonstrate the system in 5 minutes to judges without technical backgrounds. It also faced commercial scrutiny from Microsoft product leaders on the jury.

In 48 hours, I built an XGBoost demand model. It used trip history, weather, and point-of-interest density within H3 hexagonal map cells. Time features covered hour, weekday, holidays, and rolling demand. Moreover, place features covered offices, transit, and entertainment. Then Azure Maps displayed the live demand heatmap.

We competed against teams across Southeast Asia and Australia. Then we won Regional Champion across APAC. An international jury assessed the project, including Microsoft engineering and product leaders.