Why Border?

It's always annoying to spend time in traffic especially when crossing the border. It's alright. We got you! With our web-app predicting travel time, why bother to cross border?

Comment

Inspiration

Who doesn't like to drive in a highway without bothering about the traffic? I personally really like driving but hate spending time in traffic congestion. That said, with these data provided by Miovision, we want to see how well we can predict travel time especially when crossing border so that we can spend more time outside the car.

What it does

By giving our model relevant inputs such as which intersection points to start and end as well as what day and time to depart, our web-app can accurately predict the travel time for you!

How we built it

We started by cleaning the data and regroup the data into pieces that help us to build our model. We spent most of the time formulating the best strategy to tackle travel time. In the end, we came up with case-specific models that aim to provide the best solutions using xgboost algorithm characterized by the hyper-parameters. For the front-end, we want to present it in the form of web-app because it is still a prototype. We used mainly python, HTML & CSS and Javascript to build the website with flask framework as the main backbone.

Challenges we ran into

As a R-language user, the most difficult challenge for us was that we are not familiar to perform predictive analysis based on Python. It took us awhile to understand the syntax before we got the hang of it. Also, given that there are multiple datasets, it were also challenging for us to decide on how to combine variables in such a way that favors our models the most.

Accomplishments that we're proud of

We manage to create models that can accurately predict travel time when crossing border. Based on the test data, our models seem to work pretty well with the higher MSE is less than 9 minutes, given that the time interval is 15minutes.

What we learned

We learned that it is still a hurdle to move from one programming language to another even though they serves the same purpose. That said, we learned that being too comfortable with one language is not ideal so in the future, we plan to build predictive models using Python.

What's next for winhacks2020

We hope that after this event, we are still able to get more data from Miovision so that we can further our study in shortening travel time when crossing borders. Next, we also aim to look into developing mobile version of this application that can serve as a daily traffic notifier.

Share this project:

Updates