HackUNO

Team: Clash of code

Inspiration

In the midwest, oftentimes academic institutions make controversial decisions regarding whether or not to cancel school due to inclement weather. The nature of this decision is overly subjective, which inspired us to create a systematic, reliable method of determining whether or not school should be cancelled.

What it does

Whether the Weather is a software that we developed to determine whether the weather is severe enough that school should be cancelled. To do this, we analyzed historical weather data along with vehicle crash data from Des Moines, IO ranging from 2008 to 2018 to derive the severity of the road conditions during certain weather conditions.

How we built it

Much of our effort was put into ingesting the data. We found one source with historical weather data from Des Moines, and another source with vehicle crash history over the same time period. We narrowed the crash dataset based upon when poor weather conditions were evident, as well as a few more features. We merged this result with our weather dataset, and fed them to our Artificial Neural Network, which we used to predict several values based upon the weather conditions of each vehicle crash regarding the severity. For our neural network, we used the Tensor Flow library, and to parse the data we used Node js. We also used React and Js to create a front-end web application prompting the user to enter a city upon which the application predicts how many vehicle crashes there will be depending on weather data, and gives a recommendation on whether to cancel school.

Challenges we ran into

Finding and merging the dataset was a time-consuming part of our process because it was difficult to find weather and vehicle crash data that included hourly data. The hourly data was essential to give accurate predictions on how many crashes there would be. Preprocessing the data and applying the neural network to is also proved to be difficult. We spent quite a bit of time tweaking and repairing our network, because the model wasn’t converging nicely.

Team Contributions

During the entire process, we were in neat constant discussion. Many of the technologies and steps that we used for our project were new to us, so we tackled our project by pair-programming to make sure everybody got decent exposure and experience with the topics.

Built With

• css
• html
• javascript
• python
• tensorflow