I have been working on a personal project called "Predicting Stock Movement using LSTM." For my project, I have selected log returns as the input (X) and the target (y) is whether today's log returns are greater than yesterday's, represented as 0 or 1. After training and testing my LSTM model, I have achieved an overall accuracy of 70%. However, I am unsure how to obtain future predictions from it. Since the output is a binary classification, creating a future dataframe is not straightforward. Currently, the best approach I have found is to provide the model with the latest five days' log returns as inputs (using a timestep of 5), which results in a single output of either 0 or 1. This is unlike a regression problem where we can use a for loop to increment the prediction for each subsequent day. I have searched through numerous articles and watched videos on YouTube, but I have not yet found a solution. Can you help me with this issue?



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.