A Prediction of Bike Flow in Bike Renting Systems with the Tensor Model and Deep Learning

Rental bikes are popular in many urban areas to help people expand their mobility. It is important make the rental bicycle usable and available general public at appropriate time place. Inevitably, providing city with a steady supply of bicycles becomes major concern. The most aspect estimation number required each sharing station any given hour. This paper gives an examination human mobility as indicated by renting information bike system. In this paper, we proposed new approach for forecasting inflow outflow from one another during certain slots. Our method analyses pattern two steps: (1) Using Tuckers tensor decomposition create 3D model extract latent temporal spatial characteristics various stations (2) use Long-Short Term Memory Neural Network relationship between patterns derived features order predict flow stations. main contribution study that extracted through Tucker’s factorization improve accuracy prediction 16% decrease amount training data used prediction. Also, root mean squared error 1,5 bikes. We compare our baseline models historical average, ARMA, feed-forward neural network, KNN. showed best results