Answers to questions
Answerto question1: Some historical Bitcoin data can be downloaded from Kaggle, for example, https://www.kaggle.com/mczielinski/bitcoin-historical-data/data.
Once you've downloaded the dataset, try to extract the most important features and convert the dataset into a time series so that it can be fed into an LSTM model. Then the model can be trained with the time series for each time step.
Answerto question 2: Our sample project only calculates the stock price of those stocks whose actual stock price is given, and not the next day's stock price. It shows actual
and predicted
, but the next day's stock price should only contain predicted
. This is what is happening if we take predicted values as input for the next prediction:

Predicted versus actual prices for ALL
categories, where predicted values are input for the next prediction
Answerto question3: Well, then the task would be a binary classification problem. To make this happen, you need to make two changes:
- Convert the...