Reinforcement Learning Crypto Trading
Budget: €30 – €250 EUR
An algorithm for trading crypto currencies has been implemented.
Your job is to change the variables to improve the output and incorporate how to add other algorithms to the solution. We will provide you with the exact requirements that we are looking for.
Points to Note:
1. We are currently using the random slection algorithm, PPO and others like LSTM, CNN.
2. We have data from about 12 crypto coins, the main ones primarily include Bitcoin and ethereum.
3. The data ranges from 2013 upto 2020. There is no consensus currently on how much training should be used for the bot (trading agent in our case ) to mature. We have run the algo at about 200 epochs and that is sufficient to mature it.
4. The code is purely in Python
In summary: the goal is to implement reinforcement learning to make an automatic trading agent that learns through past data and make decisions to maximize the profit. The main outputs are - whether to buy, hold or sell. -The profit or loss incurred - How much coins to buy sell at which days.
The files for the code and the entire requirements are uploaded here. The Bitcoin PPO trading is the main file. The libraries needed are in the "requirements" txt file. The "model" and "utils" python files are imported into the main jupyter notebook file. The algoritm final output is the average net work, the profit gained by trading. The bitcoin_data csv file is divided into test and train sets in the python coding files.
NOTE: We are trying to explore the findings so theres no right or wrong. This is a simple project since we wont be using this for real life stock trading.
Your job is to change the variables to improve the output and incorporate how to add other algorithms to the solution. We will provide you with the exact requirements that we are looking for.
Points to Note:
1. We are currently using the random slection algorithm, PPO and others like LSTM, CNN.
2. We have data from about 12 crypto coins, the main ones primarily include Bitcoin and ethereum.
3. The data ranges from 2013 upto 2020. There is no consensus currently on how much training should be used for the bot (trading agent in our case ) to mature. We have run the algo at about 200 epochs and that is sufficient to mature it.
4. The code is purely in Python
In summary: the goal is to implement reinforcement learning to make an automatic trading agent that learns through past data and make decisions to maximize the profit. The main outputs are - whether to buy, hold or sell. -The profit or loss incurred - How much coins to buy sell at which days.
The files for the code and the entire requirements are uploaded here. The Bitcoin PPO trading is the main file. The libraries needed are in the "requirements" txt file. The "model" and "utils" python files are imported into the main jupyter notebook file. The algoritm final output is the average net work, the profit gained by trading. The bitcoin_data csv file is divided into test and train sets in the python coding files.
NOTE: We are trying to explore the findings so theres no right or wrong. This is a simple project since we wont be using this for real life stock trading.