30 thoughts on “Cryptocurrency-predicting RNN Model – Deep Learning w/ Python, TensorFlow and Keras p.11”

  1. I keep hitting this error
    train_x, train_y = preprocess_df(main_df)
    TypeError: cannot unpack non-iterable NoneType object

  2. Hi, thank you for your works and i love all of your videos

    Can you please make a video about Tensorflow Servings, that would be awesome

  3. Is there a reason to why you are selecting the number of Neurons in each layer as Exponential values of 2? eg. 32, 128?

  4. Just a small update for TF2,

    Use os.path.join instead of joining strings "/"

    log_dir = os.path.join('logs', NAME)

    tensorboard = TensorBoard(log_dir=log_dir)
    Thanks Sentdex

  5. I have just a small question that for this problem is that just want to tell whether the price will go up or go down for the future 3 more or less days for prediction, but I 'm wondered is why not to predict the future price directly by using the real time series models like RMSE evaluation optimizer?

  6. when I try to use the tensorboard page, it just sucks. It looks like a realy noob programmer made his first html page. What kind of error is this? I cant even see the graphs btw

  7. sentdex,

    checked the code five to six times already, everything should be ok, but when i try to fit the model, i get this error :

    ValueError: Error when checking input: expected cu_dnnlstm_7_input to have 3 dimensions, but got array with shape (948, 65, 1, 7)

    trying this on tensorflow 1.13.1

    please help. need your help here

  8. I get this error 'ValueError: Input 0 of layer lstm_18 is incompatible with the layer: expected ndim=3, found ndim=2. Full shape received: [None, 60]
    while running the line 'model.add(LSTM(128, input_shape=(train_x.shape[1:]), return_sequences=True))'
    Can some one pls let me know what im missing…i tried giving activation parameter too…but that dint work too….im running the CPU version…..do not have GPU

  9. if it is a binary prediction then why can't we use single output unit in the final dense layer.

  10. I ran it with exactly same parameters as your and i am getting around 99.1% accuracy in first currency run : P

  11. There is one thing I don't understand. This is based on the imported .csv data but I would like to predict the upcoming data for tomorrow for example. I could be wrong but this isn't predicting the future of a stock, it is only predicting the data from the .csv file right? How can I make an script that automatically update the csv file and predict the price of the future?
    Please let me know if I'm wrong

  12. 1 Does anyone know how I can use the model I trained with my own data and input new live data into the model through a .csv file ? (link to similar code that would work would be great to!)

    2 Which part of the code could i use to print out whether the model says it will go up or down ?

  13. Paramayning is the key advantage of P.R.I.Z.M before the rest of cryptocurrency. In the basic mechanism of Forzhinga, developers was added a unique, linear-retrograde mechanism of determination of the award for storage of funds, aimed at economic attractiveness and gradual substitution of mass of all existing Financial instruments of the world….. 1651

  14. I could get this much higher. I have some really good ideas to add to this…would make crazy profits

  15. Could you show the Checkpoints generated and how to select the "best" one to use for each currency? Thx.

  16. Hi sentdex, thank you for those videos.
    I have a question for you: If I wanted to address a regression problem (for example trying to predict the bit coin's close value and not the 0/1 target), how would I change the model in terms of layers, activation functions and metrics?

    Thank you in advance.

  17. Where you getting this error? I dont see a fix. Could that be my latest Python version? Thanks


    ValueError Traceback (most recent call last)

    <ipython-input-2-5cf4cc31bd8d> in <module>


    105 main_df['future'] = main_df[f'{RATIO_TO_PREDICT}_close'].shift(-FUTURE_PERIOD_PREDICT)

    –> 106 main_df['target'] = list(map(classify, main_df[f'{RATIO_TO_PREDICT}_close'], main_df['future']))


    108 main_df.dropna(inplace=True)

    <ipython-input-2-5cf4cc31bd8d> in classify(current, future)


    21 def classify(current, future):

    —> 22 if float(future) > float(current): # if the future price is higher than the current, that's a buy, or a 1

    23 return 1

    24 else: # otherwise… it's a 0!

    ValueError: could not convert string to float: 'Adj Close'

  18. Getting error:


    IndexError Traceback (most recent call last)

    <ipython-input-17-798aac7127a1> in <module>

    110 ## here, split away some slice of the future data from the main main_df.

    111 times = sorted(main_df.index.values)

    –> 112 last_5pct = sorted(main_df.index.values)[-int(0.05*len(times))]


    114 validation_main_df = main_df[(main_df.index >= last_5pct)]

    IndexError: list index out of range

  19. Would you please make a tutorial on how to deal with a 2D input into RNN not a flat one? Something like Convolutional LSTM. I know in Keras we have ConvLSTM2 but don't know how to apply it to a video.

  20. Hey Harrison, can you please make a part 12 video on the predict function and how to read and understand the results of that data?

Leave a Reply

Your email address will not be published. Required fields are marked *