Analysis of Tesla

Tesla is a popular company in recent years because of their innovative products,such as Model 3, Model X, Model Y, etc. The notion of Tesla is encouraging zero-carbonemission transportation and focusing on replacing fossil fuels with cleaningenergies, like electricity. So, the stock price of Tesla increasing a lot and manyinvestors care about this company. However, due to the COVID-19, the delivery ofnew vehicles is postponed, and the economy of the US is not very good. it will lead tothe decline of the stock price of Tesla and many people begin to lose confidence. In2022, the business of Tesla begins to rebound, and our team wants to create servalmodels to predict the future price of Tesla by using the historical volatility andfactors. In order to ensure the validity of our models, we would also build riskmodels: Value at Risk to measure the different risks. Therefore, we want toinvestigate the best relationship between Future Stock prices with proper risk.Based on this situation, several models are established: Model I: Multi-Factor Model;Model II: Multi-Regression Model; Model III: LSTM Model.For Model I, we try to select several features to create the multi-factor models.We choose SMB, HML, and Momentum as the three main factors to create the classicFama-French three factors model. Small Minus Big (SMB) is a size effect based onthe market capitalization of a company. High Minus Low (HML) is a value premium;it represents the spread in returns between companies with a high book-to-marketvalue ratio and companies with a low book-to-market value ratio. Momentum (Mom)is "Winner minus loser", with stocks that have outperformed in the past tend toexhibit strong returns going forward.For Model II, Time Series Model is related to predicting future outcomes,understanding past outcomes, making policy suggestions, and much more. Thesegeneral goals of time series modeling don't vary significantly from modeling cross sectionalor panel data. Time series forecasting occurs when you make scientificpredictions based on historical time stamped data. It involves building models throughhistorical analysis and using them to make observations and drive future strategicdecision-making.For Model III, long short-term memory (LSTM) normally applies recurrentneural network (RNN) as a basic recurrent unit. However, conventional LSTMassumes that the state at the current time step depends on the previous time step. Thisassumption constraints the time dependency modeling capability.For sensitivity analysis, we use the rolling mean and standard deviation of theprevious returns to generate the return of the current day and we can get thepercentage change of the rolling window, which can help us to find the sensitivity tochange in the current returns.After three models, we will make a conclusion for readers and propose our resultsby interpretations.