Multivariate Demand Forecasting for Rental Bike Systems Based on an Unobserved Component Model

Many German cities, municipalities and transport associations are expanding their bike-sharing systems (BSS) to offer citizens a cost-effective climate-friendly means of an alternative private motorized (PMT). However, operators face the challenge generating high-quality predictive analyses time series forecasts. In particular, prediction demand is key component foster data-driven decisions. To address this problem, Unobserved Component Model (UCM) has been developed predict monthly rentals BSS, whereby station-based BSS VRNnextbike, including over 2000 bikes, 297 stations 21 municipalities, employed as example. The model decomposes into trend, seasonal, cyclical, auto-regressive irregular components for statistical modeling. Additionally, includes exogenous factors such weather, user behavior (e.g., traveled distance), school holidays COVID-19 relevant covariates independent effects calculate scenario based It can be shown that UCM calculates reasonably accurate forecasts outperforms classical models ARIMA(X) or SARIMA(X). Improvements were observed in quality terms AIC/BIC (2.5% 22%) reduction error metrics from 15% 45% depending on considered model.