Skip to content

This repository provides code for training of Recurrent Neural Network for Price Prediction of Cryptocoin. Using all the data of cryptocoin, I've trained the model and got the lower RMSE score. Anyone can try out this one and apply their knowledge to get the more lower RMSE score. Refer to repository for more information regarding add-on features.

Notifications You must be signed in to change notification settings

kenil22/Cryptocoin_Future_Price_Prediciton

Repository files navigation

Cryptocoin Future Price Predicition

In this project, I am predicting future prices of the speicified bitcoin on the basis of MAX historical data. MAX refers to data which was taken since the cryptocoin came in the market. Basically, I've taken into consideration only few cryptocurrency which are mentioned in the Historical Data\sym.csv file. By leveraging the power of RNN, this is build from Tensorflow framework and using Keras API. With the minimum loss, we can train the RNN based on given data.

Data Preparatione

The yahoofinance API has been very useful for me to get real-time data of any cryptocoin.

  1. Install yahoo finance API using "pip install yfinance " .
  2. Open sym.csv file which is located in Historical Data folder. There you'll see symbol column in which you've to add symbols of crptocoins for which you want to get the real-time data.
  3. Now run Data_Preparation.py file, which will fetch the symbol of cryptocoin one by one from the csv file(refer to step 2 for csv file) . After some time, you will see csv files of the cryptocoins with real-time data in it.
  4. Now, we have all the data for training and testing.

Training

Now I've done my training on Google Colab which provides GPU which is enough to train model on this type of data. In order to train model, follow below mentioned steps:-

  1. Upload the Inferencing & Training Script\Bitcoin_Price_Prediction.ipynb on Google Colab and select GPU as runtime.
  2. From the Historical Data upload any one csv file of cryptocoin data to google colab and keep the path of that file in the training data path cell.
    Training_Path
  3. Execute remaining cells, by setting parameter values. Keep in mind while measuring RMSE score, it should be lower for better performance.
  4. Once training is completed, save the file with .h5 format and export it and save in the models directory.
  5. You can train multiple models by following step 2 to step 4.

How to run this project

Keep in mind that whatever the models you've trained, keep their symbols in the HTML file so you can get option to select symbols and view their future price. Check below attached image to get idea.

  1. create conda environment using conda create -n <your_env_name> python=3.10.10.
  2. Activate conda environment using conda activate <your_env_name>.
  3. pip install -r requirements.txt .
  4. python app.py.
Overall-View

Features

  1. We can get the data for specific period of time for selected cryptocoin.

  2. We can get the future price of selected cryptocoin.

    Future-Price-Prediction
  3. If we have any cryptocoin, let say Bitcoin. I have bought 1 Bitcoin at 5 GBP, then currently I am in Profit or Loss. You can use calculator at the top right corner for this case.

    User-Functionality
  4. We can find the correlation between some of the cryptocoins.

    Coorelation-Among-Cryptocoins

Thank you for your time !

About

This repository provides code for training of Recurrent Neural Network for Price Prediction of Cryptocoin. Using all the data of cryptocoin, I've trained the model and got the lower RMSE score. Anyone can try out this one and apply their knowledge to get the more lower RMSE score. Refer to repository for more information regarding add-on features.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published