dquant 1.2.2

DQuant is an open-source Python library for automated volatility forecasting of financial time series. It handles all stages of model construction, from raw prices to the final forecast.

DQuant

Automated Volatility Forecasting for Traders and Analysts

Volatility forecast with DQuant

---

About the Project

DQuant is an open-source Python library for automated volatility forecasting of financial time series. It handles all stages of model building: from raw prices to ready-made forecasts.

Key Idea

A trader doesn't need to know machine learning to use AI for volatility forecasting.

Features

Automated feature engineering	Creates dozens of features from raw prices (open, high, low, close, volume)
Target variable without look-ahead bias	Correct calculation of realized volatility
3 models to choose from	Gradient Boosting, XGBoost, LightGBM with early stopping
Training visualization	Error plot on train/validation to monitor overfitting
Save and load	Train once — use forever
Flexible customization	Your own features, model parameters, data sources
Integration with any data	Yahoo Finance, MetaTrader 5

Who is this for

• Algorithmic traders — for model calibration and risk management
• Discretionary traders — for assessing market regime and position sizing
• Quantitative analysts — for rapid prototyping
• Developers — for embedding into trading systems
• Students — as a ready-made benchmark and learning example

---

Installation

Requirements

• Python 3.7 or higher
• pip

pip install dquant

Verify Installation

import dquant
print(dquant.__version__)  # Should output the version

---

Quick Start

Minimal working example with Bitcoin

import pandas as pd
import yfinance as yf
from dquant.models import VolClustXGB 


# 1. Load data
df = yf.download("BTC-USD", start="2020-01-01", interval='1d')
df = pd.DataFrame({
    'open': df[('Open', 'BTC-USD')].values,
    'high': df[('High', 'BTC-USD')].values,
    'low': df[('Low', 'BTC-USD')].values,
    'close': df[('Close', 'BTC-USD')].values,
    'volume': df[('Volume', 'BTC-USD')].values
}, index=df.index)

# 2. Create model
model = VolClustXGB({}, early_stopping=True)

# 3. Train model
features = [
    'TR',
    'returns',
    'abs_returns',
    'gap',
    'body',
    'shadow',
    'close_position',
    'roll_atr_14'
]
model.fit(df, feature_list=features, input_bars=70, horizon=20, trees_count=200, show_results=True)

# 4. Make forecast
rez = model.forecast(df.iloc[-70:].copy(), show=True)

Execution result

[0.0016554 0.0018979 0.0015921 0.0014239 0.0013767 0.0011586 0.0013139
 0.0009813 0.0007931 0.0012909 0.0013664 0.0016466 0.0014836 0.0011577
 0.0008737 0.0007213 0.0008084 0.0012699 0.0015358 0.0014748]

Red shows volatility for previous candles, green shows future volatility.

---

Documentation

Resource	Description
Full documentation	All classes, methods, parameters

---

Usage Examples

With Yahoo Finance

import pandas as pd
import yfinance as yf
from dquant.models import VolClustXGB 


# 1. Load data
df = yf.download("BTC-USD", start="2020-01-01", interval='1d')
df = pd.DataFrame({
    'open': df[('Open', 'BTC-USD')].values,
    'high': df[('High', 'BTC-USD')].values,
    'low': df[('Low', 'BTC-USD')].values,
    'close': df[('Close', 'BTC-USD')].values,
    'volume': df[('Volume', 'BTC-USD')].values
}, index=df.index)

# 2. Create model
model = VolClustXGB({}, early_stopping=True)

# 3. Train model
features = [
    'TR',
    'returns',
    'abs_returns',
    'gap',
    'body',
    'shadow',
    'close_position',
    'roll_atr_14'
]
model.fit(df, feature_list=features, input_bars=70, horizon=20, trees_count=200, show_results=True)

# 4. Make forecast
rez = model.forecast(df.iloc[-70:].copy(), show=True)

With MetaTrader 5

import pandas as pd
import MetaTrader5 as mt5
import datetime as dt
from dquant.models import VolClustXGB 


symbol = "EURUSD"          # symbol to watch
timeframe = mt5.TIMEFRAME_H1   # M1, M5, M15, H1, D1, etc.
days_back = 1000             # how many days of history to load

# Connect to MT5
if not mt5.initialize():
    print("Failed to connect to MetaTrader5")
    quit()

# Check that symbol is available
if not mt5.symbol_select(symbol, True):
    print(f"Symbol {symbol} not found or not enabled")
    mt5.shutdown()
    quit()

# Calculate dates
to_date = dt.datetime.now() + dt.timedelta(hours=3)
from_date = to_date - dt.timedelta(days=days_back)

# Load bars
rates = mt5.copy_rates_range(symbol, timeframe, from_date, to_date)

mt5.shutdown()  # terminal no longer needed

if rates is None or len(rates) == 0:
    print("No data!")
    quit()

# Convert to DataFrame
df = pd.DataFrame(rates)
df['time'] = pd.to_datetime(df['time'], unit='s')

df.rename(columns={
    'tick_volume': 'volume'
}, inplace=True)

# Create model
model = VolClustXGB({}, early_stopping=True)

# Train model
features = [
    'TR',
    'returns',
    'abs_returns',
    'gap',
    'body',
    'shadow',
    'close_position',
    'roll_atr_14'
]
model.fit(df, feature_list=features, input_bars=70, horizon=20, trees_count=200, show_results=True)

# Make forecast
rez = model.forecast(df.iloc[-70:].copy(), show=True)

---

Creating an indicator for Meta Trader 5

Immediately after training the model, you can export it to a working mql5 indicator. Just one more line of code is needed:

model.save('indicator_name', type_to_save='mql5')

Done! Now you can use your trained models in Meta Trader 5.

---

How to Contribute

We welcome any contribution to the project! Here are a few ways to help:

Report a bug

Found a bug? Create an Issue with a detailed description:

• What you did
• What you expected
• What actually happened
• Code to reproduce (if possible)

Suggest an idea

Have an idea for improvement? Write to Telegram or create an Issue with the enhancement label.

---

License

The project is distributed under the MIT license. See the LICENSE file for details.

---

Project Support

If dquant has helped you in your work or studies:

• Star the project on GitHub ⭐ — it's very motivating!
• Tell your colleagues about the library
• Write to me about your experience using it

---

Contacts

Author: Denis Makarov

• Telegram: @Denchik_ai
• GitHub: @artrdon
• Project website: dquant.space