Inspiration
Food-app delivery drivers need to stop to quickly pickup/drop-off in pay parking only streets. Paying for parking cuts into their already thin bottom-line. Therefore by empowering them with more data and machine learning algorithm, they can make informed decisions! Ali was actually a pizza delivery guy and he recalls preparing a pizza slide to bribe the ticket officer during his deliveries. With RiskyTicket, perhaps he can save those slices for himself!
What it does
Using machine learning, provide an educated probability of parking ticket based on user’s geo-location. Provide hour specific probability, and nearby low risk streets.
How we built it
Machine learning algorithm was trained using past data from city of Toronto, focused DT.
DataSets Used: Toronto Parking Tickets - https://ckan0.cf.opendata.inter.prod-toronto.ca/tr/dataset/parking-tickets Traffic Volume - https://open.toronto.ca/dataset/traffic-signal-vehicle-and-pedestrian-volumes/
Trained using parking ticket dataset, and traffic volume dataset. % accuracy here:
| Model | % Accuracy | Training:Testing Ratio |
|---|---|---|
| MultipleOutputRegressor(GradientBoostingRegressor) | 93.63 | 70/30 |
| LinearSVR | 83.56 | 70/30 |
| KerasRegressor | 34.7811 | 70/30 |
| Lasso Regressor | 19.578 | 70/30 |
| Linear SVC | 8.783 | 70/30 |
| Ridge Regressor | 7.93 | 70/30 |
| Bayesian Regressor | 74.389 | 70/30 |
| DecisionTreeRegressor | 23.58 | 70/30 |
| LSTM(TANH) | 42.17 | 70/30 |
| LSTM(RELU) | 38.98 | 70/30 |
| RepeatedKFold | 0.128 | 70/30 |
| kNNRegressor | 1.365 | 70/30 |
High traffic areas with high ticket counts = High Risk Low traffic areas with low ticket counts = Low Risk
Challenges we ran into
Finding the right dataset Scaling the data to run on machines
What's next for RiskyTicket
Build a more robust web application around the model. As of now, there is limited functionality (and lots of bugs!) Once there is a better idea of deliverables, move to the more convenient mobile platform. We want to build a system that food-app delivery drivers can use to make informed decisions when deciding whether to pay for parking.
Log in or sign up for Devpost to join the conversation.