Prediction of Open Slots in Bicycle Parking Stations Using the Decision Tree Method

Most developed countries are now emphasizing the use of green energy and are less reliant upon fossil fuels to avoid the greenhouse effect. Governments and societies highly recommend the use of a bicycle for green energy. One of the issues faced regarding these initiatives is finding an available parking slot near the targeted station for private or shared bicycles. Several modern technology solutions, like IoT, GPS devices, and ML prediction techniques, were used to help the public. In this paper, one of the prediction method techniques discussed is the Decision Tree algorithm, which analyzes past events and predicts the future so the user can change the time to park the bicycle or switch to another parking lot, which would be more feasible and save time in finding a free parking space. The dataset was taken from the SF Bay Area bike share, which contained every dock station's ID, timestamp, and how many bicycles and free slots were available at the given time. The 80/20 rule has been used to train and test the dataset training and testing. An increase in accuracy was observed from 47% to 93% after reducing the dataset size from 100% to 20%.