CryptocurrenciesPricePrediction

A Neural Network LSTM model that predicts price of cryptocurrencies.

Coded by Caner Sülüşoğlu and Metin Cem Demirdaş

Setup Environment

To create a python virtual environment.

python -m venv venv or python3 -m venv venv

Activate your virtual env.

.\venv\Scripts\activate

Installing requirements .

pip install -r ./requirements.txt

Starting train the model.

python ./main.py

Arguments Informations

Using -h or --help keyword, informations about arguments will be shown.

Train Arguments Table

These parameters are using with train keyword:

Short	Long	Default	Description
-c	--currency	BTC-USD	Crypto currency parity.
-d	--download_dataset	False	Download current time dataset?
-udd	--use_downloaded_dataset	False	Use already downloaded dataset?
-lm	--load_model	False	Load trained model?
-e	--epoch	200	Epoch number.
-l	--lookback	60	Subset count that is looking backward.
-f	--forecast	30	Days count that will be predicted in future.

Example: python train -c 'BTC-USD' -d -udd -e 100 -l 120 -f 60

Test Arguments Table

These parameters are using with test keyword:

Short	Long	Default	Description
-c	--currency	BTC-USD	Crypto currency parity.
-l	--lookback	60	Subset count that is looking backward.
-f	--forecast	30	Days count that will be predicted in future.

Example: python test -c 'BTC-USD' -l 120 -f 60

Pre Trained Models Screenshots

Currency = BTC-USD

Epcoh = 200

Batch Size = 50

Lookback	Forecast	Loss	All Time
60	30
180	90
360	180
720	360