The development of the new energy automobile industry has attracted the attention of many investors and seniors, with the stock price fluctuations of BYD, Chana Auto and TESLA, as representatives of leading enterprises, being a particular focus. This paper uses the forecasting method of time series analysis to fit and forecast the daily logarithmic revenue series of three new energy companies. The ARIMA, GARCH and LSTM models were each constructed for empirical analysis. The accuracy of the model fitting was assessed by comparing the relative gap between the actual and predicted value, and MSFE and MAFE were used as indicators to compare the final models. The final results show that the GARCH and LSTM models outperform the ARIMA model in capturing the volatility clustering effect and eliminating residual autocorrelation for BYD and Chana, while all three models perform well for Tesla. In terms of predictive accuracy, the LSTM model excels in handling BYD stock data, while the ARIMA and GARCH models show advantages in predicting Tesla and Chana stock returns. In essence, this paper presents a preliminary examination of the predictive efficacy of three distinct model types applied to the daily logarithmic return series of stocks. Each model exhibits a unique aptitude for addressing specific characteristics of this series. This encourages researchers to employ multiple models in conjunction with one another to ascertain a superior model for prediction in a given series.