<p>The pandemic emergencies such as COVID-19 can cause unprecedented disruptions in global supply chains, and make the sales of manufacturing products become rather difficult to forecast since the historical data of manufacturing companies may have acute and wide-spread biases under such new situations, and the traditional forecasting models become no longer valid during the pandemic. In this study, we collaborate with two automobile manufacturers to assemble a unique proprietary operational dataset, and implement a variety of methods to forecast the sales of cars during the COVID-19 pandemic. In particular, enormous evidence has shown that consumer behavior can be significantly affected by social media information during the pandemic, we conduct extensive data mining analysis to identify how consumer behavior changes for Chinese automobile industry, and integrate social media information into our forecasting models. We find that our proposed boosting method performs significantly better than traditional forecasting models and other classical machine learning methods. We also observe that the results could still be improved when the social media information are incorporated. With our analysis and results, firms may adopt useful forecasting models to quickly anticipate the shift in consumer behavior and adjust their production planning accordingly. We also outline how policymakers can implement policies to improve automobile supply chain continuity in the context of pandemic emergencies.</p>