The CitiBike Demand Prediction Project aims to address the challenge of optimizing bike supply for daily rentals, particularly in urban areas with a focus on Oakland and San Francisco. The motivation for this project arises from the need to better manage bike sharing services, considering factors such as weather, seasonal patterns, and demographic information. The primary stakeholders are bike-sharing businesses looking to enhance operational efficiency and cities struggling with issues like homelessness and bike theft.
The existing data available for bike-sharing services lacks essential information, such as weather conditions and demographic patterns, making it challenging to accurately predict the demand for bikes on a daily basis. This deficiency leads to problems like oversupply or undersupply of bikes, impacting the overall success of the business. Additionally, addressing the homeless problem and bike theft is crucial for the sustainable growth of bike-sharing services in cities like Oakland and San Francisco.
Machine Learning (ML) provides an ideal approach to solving this problem due to its ability to analyze large datasets, identify patterns, and make predictions based on historical data. The use of time series analysis allows us to make daily predictions, while clustering helps uncover seasonal patterns in bike demand and supply. By employing classification models, we aim to identify specific demographics and user behaviors that contribute to daily bike demand. The focus on interpretability ensures that the models provide insights that are actionable for business decisions.
-Clean and perform extensive data analysis on Citi Bike Data from UC Irvine Academic Data Set spanning over 2 years -Develop predictive models using time series analysis, clustering, and classification to accurately forecast daily bike demand. -Prioritize interpretability over raw accuracy to facilitate informed decision-making by bike-sharing businesses. -Implement ARIMA and LSTM models for series forecasting to enhance prediction accuracy. -Expand the bike rental business to other cities and regions by providing reliable demand and supply predictions. -Address homelessness and bike theft issues in Oakland and San Francisco through better management of bike-sharing services.
-Linear Regression -LSTM (Long-Short-Term-Memory) Neural Network -ARIMA -Random Forest