Paper Title
UTILISING TRANSFER LEARNING AND GAN FOR STOCK MARKET DATA ANALYSIS
Abstract
Abstract - In order to Analyzestock market data, this research investigates the use of Generative Adversarial Networks (GANs) in conjunction with transfer learning. The two main stages of the experiment are as follows: first, synthetic datasets are created using an LSTM-based generator and an MLP-based discriminator that have been trained using historical Tesla stock market data that was collected from Yahoo Finance. Second, a multi-layered LSTM model is created and trained on a larger synthetic dataset first, then it is transferred to the original real dataset for transfer learning. Remarkably, without any transfer learning, the model's performance peaked when it was applied straight to the real dataset. Nonetheless, when the model was first trained on artificial datasets and subsequently refined by transfer learning, nearly identical performance levels were attained. Interestingly, the research shows that expanding the artificial dataset's size does not consistently improve the model’s performance
Keywords - GAN, LSTM, Transfer Learning