Skip to content

Commit

Permalink
Currency Arbitrage with Reinforcement Learning added
Browse files Browse the repository at this point in the history
  • Loading branch information
@ashis2004
ashis2004 committed Jul 19, 2024
1 parent d414024 commit cda1f1e
Show file tree
Hide file tree
Showing 3 changed files with 686 additions and 0 deletions.
290 changes: 290 additions & 0 deletions ... Arbitrage with Reinforcement Learning/agent_for_currency_arbitrage_using_Rlearning.ipynb
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,290 @@
{
"nbformat": 4,
"nbformat_minor": 0,
"metadata": {
"colab": {
"provenance": []
},
"kernelspec": {
"name": "python3",
"display_name": "Python 3"
},
"language_info": {
"name": "python"
}
},
"cells": [
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"id": "ARmpyIVdteu5"
},
"outputs": [],
"source": [
"import pandas as pd\n",
"import numpy as np\n",
"import random\n",
"from datetime import datetime, timedelta\n",
"\n",
"# Parameters\n",
"num_entries = 1000\n",
"currency_pairs = ['EUR/USD', 'USD/JPY', 'GBP/USD', 'USD/CHF', 'AUD/USD', 'USD/CAD']\n",
"\n",
"# Generate timestamps\n",
"start_time = datetime.now()\n",
"timestamps = [start_time - timedelta(minutes=5*i) for i in range(num_entries)]\n",
"\n",
"# Generate synthetic data\n",
"data = []\n",
"for i in range(num_entries):\n",
" timestamp = timestamps[i]\n",
" pair = random.choice(currency_pairs)\n",
" bid_price = round(random.uniform(1.0, 1.5), 4)\n",
" ask_price = round(bid_price + random.uniform(0.0001, 0.0005), 4)\n",
" data.append([timestamp, pair, bid_price, ask_price])\n",
"\n",
"# Create DataFrame\n",
"columns = ['Timestamp', 'Currency Pair', 'Bid Price', 'Ask Price']\n",
"df = pd.DataFrame(data, columns=columns)\n",
"\n",
"# Save to CSV\n",
"df.to_csv('currency_arbitrage_data.csv', index=False)\n"
]
},
{
"cell_type": "code",
"source": [
"import pandas as pd\n",
"\n",
"# Load the dataset\n",
"df = pd.read_csv('currency_arbitrage_data.csv')\n",
"\n",
"# Convert the Timestamp column to datetime\n",
"df['Timestamp'] = pd.to_datetime(df['Timestamp'])\n",
"\n",
"# Display the first few rows of the dataset\n",
"print(df.head())\n"
],
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "tUTB5cjltxSh",
"outputId": "fd4c12e8-afb0-423c-a5ad-634b96b2bdcf"
},
"execution_count": 2,
"outputs": [
{
"output_type": "stream",
"name": "stdout",
"text": [
" Timestamp Currency Pair Bid Price Ask Price\n",
"0 2024-07-19 11:38:48.716950 USD/CHF 1.2220 1.2224\n",
"1 2024-07-19 11:33:48.716950 USD/JPY 1.4628 1.4631\n",
"2 2024-07-19 11:28:48.716950 USD/JPY 1.3268 1.3271\n",
"3 2024-07-19 11:23:48.716950 USD/CHF 1.1257 1.1262\n",
"4 2024-07-19 11:18:48.716950 USD/CAD 1.3318 1.3319\n"
]
}
]
},
{
"cell_type": "code",
"source": [
"from sklearn.preprocessing import StandardScaler, LabelEncoder\n",
"\n",
"# Encode the currency pairs\n",
"le = LabelEncoder()\n",
"df['Currency Pair'] = le.fit_transform(df['Currency Pair'])\n",
"\n",
"# Normalize the bid and ask prices\n",
"scaler = StandardScaler()\n",
"df[['Bid Price', 'Ask Price']] = scaler.fit_transform(df[['Bid Price', 'Ask Price']])\n",
"\n",
"print(df.head())\n"
],
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "krofbKLUtxPK",
"outputId": "006a9e1e-d992-4c06-e43e-70763da549c6"
},
"execution_count": 3,
"outputs": [
{
"output_type": "stream",
"name": "stdout",
"text": [
" Timestamp Currency Pair Bid Price Ask Price\n",
"0 2024-07-19 11:38:48.716950 4 -0.173319 -0.172652\n",
"1 2024-07-19 11:33:48.716950 5 1.508608 1.508600\n",
"2 2024-07-19 11:28:48.716950 5 0.558683 0.558661\n",
"3 2024-07-19 11:23:48.716950 4 -0.845950 -0.844594\n",
"4 2024-07-19 11:18:48.716950 3 0.593606 0.592188\n"
]
}
]
},
{
"cell_type": "code",
"source": [
"import numpy as np\n",
"import gym\n",
"from gym import spaces\n",
"\n",
"class CurrencyArbitrageEnv(gym.Env):\n",
" def __init__(self, df):\n",
" super(CurrencyArbitrageEnv, self).__init__()\n",
" self.df = df\n",
" self.current_step = 0\n",
"\n",
" # Actions: 0 = hold, 1 = buy, 2 = sell\n",
" self.action_space = spaces.Discrete(3)\n",
"\n",
" # Observation space: bid price, ask price, and currency pair\n",
" self.observation_space = spaces.Box(low=-np.inf, high=np.inf, shape=(3,), dtype=np.float32)\n",
"\n",
" # Initial account balance\n",
" self.balance = 10000\n",
"\n",
" def reset(self):\n",
" self.current_step = 0\n",
" self.balance = 10000\n",
" return self._next_observation()\n",
"\n",
" def _next_observation(self):\n",
" obs = self.df.iloc[self.current_step][['Bid Price', 'Ask Price', 'Currency Pair']].values\n",
" return obs\n",
"\n",
" def step(self, action):\n",
" obs = self._next_observation()\n",
" self.current_step += 1\n",
"\n",
" if action == 1: # Buy\n",
" self.balance -= obs[1] # Subtract ask price\n",
" elif action == 2: # Sell\n",
" self.balance += obs[0] # Add bid price\n",
"\n",
" reward = self.balance\n",
"\n",
" done = self.current_step >= len(self.df) - 1\n",
"\n",
" return obs, reward, done, {}\n",
"\n",
" def render(self, mode='human'):\n",
" print(f'Step: {self.current_step}, Balance: {self.balance}')\n",
"\n",
"# Create the environment\n",
"env = CurrencyArbitrageEnv(df)\n"
],
"metadata": {
"id": "pzTAsCD3txKv"
},
"execution_count": 4,
"outputs": []
},
{
"cell_type": "code",
"source": [
"import random\n",
"from collections import deque\n",
"\n",
"class QLearningAgent:\n",
" def __init__(self, state_size, action_size):\n",
" self.state_size = state_size\n",
" self.action_size = action_size\n",
" self.memory = deque(maxlen=2000)\n",
" self.gamma = 0.95 # discount rate\n",
" self.epsilon = 1.0 # exploration rate\n",
" self.epsilon_min = 0.01\n",
" self.epsilon_decay = 0.995\n",
" self.learning_rate = 0.001\n",
" self.q_table = {}\n",
"\n",
" def get_qs(self, state):\n",
" return self.q_table.get(tuple(state.flatten()), np.zeros(self.action_size))\n",
"\n",
" def remember(self, state, action, reward, next_state, done):\n",
" self.memory.append((state, action, reward, next_state, done))\n",
"\n",
" def act(self, state):\n",
" if np.random.rand() <= self.epsilon:\n",
" return random.randrange(self.action_size)\n",
" qs = self.get_qs(state)\n",
" return np.argmax(qs)\n",
"\n",
" def replay(self, batch_size):\n",
" minibatch = random.sample(self.memory, batch_size)\n",
" for state, action, reward, next_state, done in minibatch:\n",
" target = reward\n",
" if not done:\n",
" target = reward + self.gamma * np.amax(self.get_qs(next_state))\n",
" qs = self.get_qs(state)\n",
" qs[action] = target\n",
" self.q_table[tuple(state.flatten())] = qs\n",
" if self.epsilon > self.epsilon_min:\n",
" self.epsilon *= self.epsilon_decay\n",
"\n",
"# Initialize the agent\n",
"state_size = env.observation_space.shape[0]\n",
"action_size = env.action_space.n\n",
"agent = QLearningAgent(state_size, action_size)\n",
"\n",
"# Train the agent\n",
"episodes = 300\n",
"batch_size = 32\n",
"\n",
"for e in range(episodes):\n",
" state = env.reset()\n",
" state = np.reshape(state, [1, state_size])\n",
"\n",
" for time in range(50):\n",
" action = agent.act(state)\n",
" next_state, reward, done, _ = env.step(action)\n",
" next_state = np.reshape(next_state, [1, state_size])\n",
"\n",
" agent.remember(state, action, reward, next_state, done)\n",
" state = next_state\n",
"\n",
" if done:\n",
" print(f\"Episode: {e+1}/{episodes}, Score: {reward}, Epsilon: {agent.epsilon:.2}\")\n",
" break\n",
"\n",
" if len(agent.memory) > batch_size:\n",
" agent.replay(batch_size)\n"
],
"metadata": {
"id": "l7MHaOXjtxGQ"
},
"execution_count": 6,
"outputs": []
},
{
"cell_type": "code",
"source": [
"# Evaluate the agent\n",
"state = env.reset()\n",
"state = np.reshape(state, [1, state_size])\n",
"total_reward = 0\n",
"\n",
"for time in range(500):\n",
" action = agent.act(state)\n",
" next_state, reward, done, _ = env.step(action)\n",
" next_state = np.reshape(next_state, [1, state_size])\n",
" total_reward += reward\n",
" state = next_state\n",
"\n",
" if done:\n",
" print(f\"Final Balance: {total_reward}\")\n",
" break\n"
],
"metadata": {
"id": "t_-a_DbRuDay"
},
"execution_count": 7,
"outputs": []
}
]
}
Loading

0 comments on commit cda1f1e

Please sign in to comment.