(Jul 11, 2021 09:55 PM)m8than Wrote: This is called confirmation bias. Your model is probably correct by mostly luck but you want to believe it actually works even though it doesn't work for the task you set it (predicting prices)
No this is called LSTM Time-series machine learning where a computer learns to predict point in a time-series sequence like price movement. I have revised the model and will add new predictions soon.
(Jul 11, 2021 09:55 PM)m8than Wrote: This is called confirmation bias. Your model is probably correct by mostly luck but you want to believe it actually works even though it doesn't work for the task you set it (predicting prices)
The model results in a 60% accuracy of replicating bitcoin movement. You dont 'believe' anything when training a model. You train it and look at the results.
Update
The original predictions were:
2021-07-10 to 2021-07-11 Market direction:
DOWN
2021-07-11 to 2021-07-12 Market direction
UP
2021-07-12 to 2021-07-13 Market direction
DOWN
Between 2021-07-10 and 2021-07-11 the market dropped 3.7%.
Between 2021-07-11 and 2021-07-12 the market went 3.6% bullish.
Between 2021-07-12 and creation of post the market is down 4.9%.
In conclusion the model has been able to predict market direction but there is still uncertainty whether this was a coincidence or not.
New Predictions
2021-07-13 to 2021-07-14
DOWN
2021-07-14 to 2021-07-15
UP