ocean_protocol_eth_pred

Predict Eth price

Install

with ocean-lib and data science tools + jupyter into vscode

conda clean -p pip cache purge conda update -n base -c conda-forge conda conda create --name ocean_tf python=3.10.4 conda activate ocean_tf conda install -c apple tensorflow-deps==2.9.0 python -m pip install tensorflow-macos==2.9.0 python -m pip install tensorflow-metal==0.5.0

pip3 install matplotlib ccxt eth_account conda install pandas conda install scikit-learn

ARCHFLAGS="-arch arm64" pip3 install ocean-lib

pip install ipykernel==6.21.0 --force-reinstall

ETH Prediction Round 3 : 1h timeframe + report:

• 03_Data_Sources.ipynb : Data pipeline
• 04_Explore.ipynb : Explorartion + Model
• /doc/ETH Prediction Round 3 - jeugregg.pdf : final report

ETH Prediction Round 4 : 1 hour timeframe:

• 05_Explore_april
• 06_Model

ETH Prediction Round 5 : 5 min timeframe:

• 07_Data_Sources_05min ( limited to ohlcv data)
• 06_Model_05min (Auto-encoder + LSTM)
• Readme - https://github.com/oceanprotocol/predict-eth/blob/main/challenges/main5.md
• Results / Prizes - 1st place : https://blog.oceanprotocol.com/here-are-the-winners-of-the-predict-eth-round-5-data-challenge-95f71bcade95

ETH Prediction Round 6 : 5 min timeframe:

• 08_Data_Sources_05min (with all data available)
• 09_Model_05min (Auto-encoder + LSTM) - Models not optimized (10 epochs max)
• Blog
  - https://blog.oceanprotocol.com/predict-eth-round-6-data-challenge-is-live-d305502888f9
• Readme
  - https://github.com/oceanprotocol/predict-eth/blob/main/challenges/main6.md
• Final Transaction
  - https://mumbai.polygonscan.com/tx/0x32ddbba3c3f4fc2664570b7f72fce47cbf29f64318a5e445bf8debd917618418
• Results / Prizes - 2nd place / 15 : https://medium.com/oceanprotocol/here-are-the-winners-of-the-predict-eth-round-6-data-challenge-9b8e8f786170

ETH Prediction Round 7 : 5 min timeframe:

• 10_Data_Sources_05min (with all data available)
• 11_Model_05min (Auto-encoder + LSTM) - Models with auto-select best encoded features
• Blog
  - https://blog.oceanprotocol.com/here-are-the-winners-of-the-predict-eth-round-7-data-challenge-d5ec1c2056e0
• Readme
  - https://github.com/oceanprotocol/predict-eth/blob/main/challenges/main7.md
• Final Transaction
  - https://mumbai.polygonscan.com/tx/0xa5c13583cded7e69140ae04e4730837a5e8e97ae4decc7dc9a921e388ca71a39
• Results / Prizes - 4th place / 80 : https://medium.com/oceanprotocol/predict-eth-round-7-data-challenge-is-live-9be6d7faa4a1

2023/08/02 : ETH Prediction Round weely : 5 min timeframe:

• 10_Data_Sources_05min (with all data available)
• 12_Model_05min (Auto-encoder 2 LSTM layers + LSTM predictor)
• Blog
  - https://medium.com/oceanprotocol/introducing-challenge-data-farming-378bba28fc97
• Readme
  - https://github.com/oceanprotocol/predict-eth/blob/main/challenges/challenge-df.md
• Final Transaction
  - https://mumbai.polygonscan.com/tx/0xc2f8a73272a69e81db02149a21d0ff0899acecbe53077642509cec373175e791

2023/08/09 : round

• 13_Model_05min.ipynb
  • Test with 6h of lag
  • TRAIN/TEST : 80/20
  • Model without dropout for decoder part of AE

2023/08/16 : round

• 14_Model_05min.ipynb
  • Test with 6 hours of lag
  • TRAIN/TEST : 80/20
  • remove some feature : sin/cos month
  • normalized by ETH price some economics indices : spx, fvx, dxy because TEST data out of TRAIN range
  • Very slow to train on TRAIN set
  • MSE TEST lower than MSE TRAIN at first

2023/08/30 & 2023/09/06: round 52 & 53

• 16_Model_05min.ipynb
  • Test with 6 hours of lag
  • TRAIN/TEST : 80/20
  • remove some feature : sin/cos month
  • normalized by ETH price some economics indices : spx, fvx, dxy because TEST data out of TRAIN range