With the advancement of deep learning technologies, new methods and perspectives have emerged for forecasting stock market trends. However, there is a relative lack of evaluative research on stock trend prediction. Most existing literature is predominantly review-based, offering textual summaries of AI models without detailed empirical analysis. Furthermore, studies on stock trend prediction often emphasize singular experimental validations, focusing primarily on horizontal comparisons among similar models while neglecting comprehensive, multidimensional evaluations. In this study, we analyze the components of the CSI300 and CSI500 indices from the A-share market, collecting extensive datasets spanning a significant time period. To extract features from this data, we employ the Alpha158 and Alpha360 factors. We then conduct empirical testing using eight distinct deep learning algorithms. By evaluating the models across nine performance metrics and three loss calculation indicators, we rigorously validate their effectiveness. This research not only establishes a multidimensional comparative framework but also provides a more thorough and nuanced analytical perspective on stock market prediction. It aims to contribute to the advancement of this field and to promote further development in the application of deep learning to financial forecasting.