Leveraging AI-Based Approaches to Forecast Bike Demand in Smart Citites

Owing to global environmental limitations, urban areas, especially in terms of urban transportation, need to prioritize eco-friendly solutions. Sustainable Bike-Sharing Systems (BSS) have emerged as a significant component of the global transportation network, offering an alternative to gaspowered vehicles in cities. However, the placement of docking stations relies on factors such as proximity to cycling paths, density of the population, ease of bike accessibility, commuting expenses, the distribution of docks, and route imbalances. This research aims to assess and compare the efficacy of various machine learning and time series algorithms in forecasting bicycle demand based on historical data. Specifically, the study tested RF, ARIMA, and SARIMA algorithms, evaluating their performance using metrics such as MSE, MAE, and RMSE. The study revealed that the random forest (RF) algorithm showed superior performance compared to the time series models, whereas ARIMA and SARIMA demonstrated less accuracy, likely due to their assumptions of linearity and stationarity within the dataset. Overall, the research offers valuable insights into the effectiveness of different algorithms for predicting bike demand, laying the groundwork for further exploration in this area.