juvaana/TANZANIA-S-URBAN-MOBILITY-CHALLENGE

Predict Peak Daladala Demand. Build Smarter Cities. Tanzania’s cities move fast. Yet transport demand remains unpredictable. With this challenge. Your mission is simple & powerful, build a machine learning model that predicts when a daladala route will hit peak demand. You will be working with a simulation data built at national scale. # TANZANIA'S URBAN MOBILITY CHALLENGE 🚍 ## Predict Peak Daladala Demand. Build Smarter Cities. Predict Peak Daladala Demand. Build Smarter Cities. This challenge asks you to build a machine learning model that predicts when a daladala route will reach peak demand. This is not theoretical and not toy data. You will work with a national-scale urban mobility simulation dataset. Welcome to Juvaana. ## Why This Matters 📊 Urban transport inefficiency affects productivity and quality of life. Several factors influence mobility demand: Rush hour congestion costs time and economic output ,Rainfall changes commuter behavior,End-of-month salary cycles alter transport demand ,Population density drives route pressure If we can accurately predict peak demand: Operators can optimize dispatch , Cities can reduce congestion,Transportation systems become smarter and data-driven This is how data transforms infrastructure. ## The Challenge You are provided with historical urban mobility simulation data from five Tanzanian cities: Dar es Salaam ,Mwanza,Arusha,Dodoma and Mbeya Each row represents a daladala route at a specific hour. ## Your task: High demand → peak = 1 Normal demand → peak = 0 The concept is simple. The modeling challenge is serious. ## The Data Participants receive two files: train.csv → Features + target (peak) test.csv → Features only Important signals in the dataset include: Morning rush hours (6–9 AM) ,Evening rush hours (4–8 PM) Higher demand intensity in Dar es Salaam ,Weather influencing commuter patterns Salary cycles affecting travel behavior ,Population density effects Real-world characteristics: Distribution shift between train and test data Noisy variables included Imperfect and messy data (just like real urban systems) ## Evaluation Metric Submissions are evaluated using the F1 Score. F1 balances: Precision Recall This prevents models from simply predicting the majority class. Higher F1 Score → Better mod …