The Project

The purpose of our weather forecasting application was to create a "down to the minute" prediction service for Android phones. To this end, we created hourly predictions of up to 6 hours. We were thinking of an "on the go" audience who needed a quick, no nonsense application. Something that they could look at quickly and continue on with their busy lives

Like most college students, we didn't have access to supercomputers or great sums of resources to find and create the most accurate create our predictions, thus we used four different approaches to try to create as accurate a prediction as possible: Evolutionary Computing, Image Processing, Machine Learning and Modelling. These four approaches had different strengths and weaknesses which we used to predict various types of weather, from temperature and wind to the movement of storm cells. On their own, each method was weak, but together they formed a relatively accurate picture of weather patterns in the Minneapolis area. We were also able to create weather alerts, just in case the user wanted to be warned of any sort of weather, either in the present time or ahead of time.

In an ideal world, we could've expanded further with our project. After our successful run in Minneapolis, we ostensibly could have expanded to nationwide predictions. We also hoped to integrate this application into a scheduling application, so extreme weather alerts would be automatically generated if it could interfere with one's plans.