What is it?

A physical model to map the predicted bike station traffic over the University of Maryland

​

How does it work?

First we manipulated the dataset by extracting  the date and time separately and converting them into day of the week and 24 hour time format respectively. For feeding the data into our machine learning model we labeled the gender, station and location with numeric value. For prediction system we built a 5 hidden-layer deep neural network containing 4 hidden units in each layer. We aggregated the predicted cycle demand for each hour for each day at each location and station. Whenever a cycle is borrowed from that station the count is reduced. This entire data is visualized physically using Arduino controller. A map is created for University of Maryland and designated stations are highlighted. For each station the number of cycles currently at station are indicated with blinking leds. The number of LEDs blinking proportionately indicate the number of cycles at the stand

🏆🏆 2018 Data Challenge Grand Prize Winner
🏆🏆 2018 EqualsEquals Hackathon 2nd Place Winner