I liked the concept and wanted to extend the training dataset ... so I pulled the data for one more month and then moved the resulting file from the "raw_historic_data" to the datasets/BTC-USD folder.

Then changed the config.py to extend it by one more month, like this:

TRAINING_MONTHS = ["2018_06","2018_07","2018_08","2018_09","2018_10","2018_11","2018_12","2019_01",
                   "2019_02","2019_03","2019_04","2019_05","2019_06","2019_07","2019_08","2019_09",
                   "2019_10","2019_11","2019_12","2020_01","2020_02","2020_03","2020_04","2020_05",
                    "2020_06","2020_07","2020_08","2020_09","2020_10"]

TESTING_MONTHS = ["2020_11"]

..... but I'm getting the following "type mismatch" error:

Traceback (most recent call last):
  File "c:\Development\crypto-trading-ai-bot-basic\Controller.py", line 68, in <module>
    x_test, y_test, prices = dataset.createTrainTestSets(COIN_PAIR, data, training_window=TRAINING_WINDOW, labeling_window=LABELING_WINDOW)
  File "c:\Development\crypto-trading-ai-bot-basic\Dataset.py", line 169, in createTrainTestSets
    change_rate = (next_close_price/latest_close_price) - 1
TypeError: unsupported operand type(s) for /: 'str' and 'float'

I think you could add indicators' historical data to improve bot predictions such as MACD, RSI, EMAs, etc.

It would be very useful to have a requirements.txt file listing required package versions. I am currently seeing a deadlock in package versions:

ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
cbpro 1.1.4 requires six==1.10.0, but you have six 1.16.0 which is incompatible.

ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
tensorflow 2.9.1 requires six>=1.12.0, but you have six 1.10.0 which is incompatible.

Why are the data arrays shuffled on this line?

This results in the buy and sell prices being random when running the bot. This means the results are completely random.

You will see what I am talking about when you change the following lines:

            print(">> BUYING $",self.trade_amount," WORTH OF BITCOIN FOR", self.account.btc_price)

and

            print(">> SELLING $",self.trade_amount," WORTH OF BITCOIN FOR", self.account.btc_price)

Please see the following Traceback:

> Trading Automatically for ['2020_10']
#################################################################################################

Account Balance: $ 1000 BTC: $ 0 USD: $ 1000

#################################################################################################
########################################## DAY 1 #########################################
Traceback (most recent call last):
File "Controller.py", line 75, in
auto_trader.runSimulation(x_test, prices)
File "/home/crypto-trading-ai-bot-basic/AutoTrader.py", line 54, in runSimulation
prediction = self.advisor.predict(np.array([samples[i]]))
File "/home/crypto-trading-ai-bot-basic/Model.py", line 73, in predict
prediction = np.array(tf.argmax(self.model.predict(sample),1))[0]
IndexError: too many indices for array: array is 0-dimensional, but 1 were indexed

Any idea what could be wrong?

i am not sure how you come up with data set ? or how to collect it to apply to a different kinds of trade or more months.

I know your code is run base on historical data for stimulation only. How do we convert it to run real-time data to evaluate how good or bad it is?

= check every 30 sec or 60 sec
= do buy or sell on the virtual account

I see you have no LICENSE file for this project. The default is copyright.

I would suggest releasing the code under the AGPL-3.0-or-later license so that others are encouraged to contribute changes back to your project.