/BTC-Prediction-by-LSTM

Using LSTM model to predict Bitcoin prices. Our project is chosen to be One Of Outstanding Projects that are qualified to present in front of reviewers from industry partners of NUS MSBA and the whole class

Primary LanguageJupyter Notebook

BTC Prediction by LSTM

The project mainly try to use LSTM and static database to predict BTC prices.

  • Time Window: 10 days
  • predict horizon: 1 day average price
  • Features in total: 61
  • Dated Between：2020/9/16 ~ 2021/9/17

The main directory and files of this code are as follows:
--------- cleaned data_new.csv (Raw cleaned data)
--------- all_data_new.csv (Raw integrated data)
--------- bitcoin_final.ipynb (Code)
--------- Bitcoin price prediction.pdf (Report)

1 Data Input

1.1 Bitcoin property and network

Bitcoin Daily Average Prices, Hash Rate, Miner Rewards, Miner Reserves,…

1.2 Bitcoin Marketing and trading

Number of Large Transactions, Average Transaction Size, Average Balance, Average Time Between Transactions, …

1.3 Global economic indicators

Gold price, US dollar index, Dow Jones Commodity index, …

1.4 Investors and Media Attention

Google trend, Twitter positive, Twitter negative, …

1.5 Prices of Other Cryptocurrencies and BTC Index

Ethereum, Dogecoin, CCI30*

2 Model

3 Model Improvement

Steps RMSE
Original Attempt 0.04543
Adding Features 0.03101
Feature Selection 0.0264
Smoothing 0.0264
Final RMSE 0.02631

Specially thanks for Prof. Pang and other indutry partners of MSBA for their guidence and review.

Our project is chosen to be One Of Outstanding Projects that are qualified to present in front of reviewers from industry partners of NUS MSBA and the whole class.