As the world's largest auto producer and consumer, China is both the most promising and complex market given the country's rapid economic growth, huge population, and many regional and segment preference differences. This research is aimed at developing data-driven demand models for customer preference analysis and prediction under a competitive market environment. Regional analysis is first used to understand the impact of geographical factors on customer preference. After a comprehensive data exploration, a customer-level mixed logit model is built to shed light on fast-growing vehicle segments in the Chinese auto market. By combining the data of vehicle purchase, consideration, and past choice, cross-shopping behaviors and brand influence are explicitly modeled in addition to the impact of customer demographics, usage behaviors, and attributes of vehicles. Scenario analyses are performed for segment demand forecasting by examining influencing factors such as economic change, fuel economy improvement and infrastructure development. Finally, a new network analysis approach is proposed to model customer cross-shopping behaviors that can inform the firm about the implied market structure and product competitive positioning. Our proposed approach is demonstrated by using a rich set of market data collected in China.