{
"cells": [
{
"cell_type": "code",
"execution_count": 2,
"id": "0ba75c87",
"metadata": {},
"outputs": [],
"source": [
"# Data and analysis libraries\n",
"import polars as pl # Fast dataframes for financial data\n",
"import numpy as np # Numerical computing library\n",
"from datetime import datetime, timedelta # Date and time operations\n",
"import random\n",
"\n",
"# Machine learning libraries \n",
"import torch # PyTorch framework\n",
"import torch.nn as nn # Neural network modules\n",
"import torch.optim as optim # Optimization algorithms\n",
"import research # Model building and training utilities\n",
"\n",
"# Visualization and \n",
"import altair as alt # Interactive visualization library\n",
"\n",
"# data sources\n",
"import binance # Binance market data utilities"
]
},
{
"cell_type": "code",
"execution_count": 8,
"id": "ee4d8884",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"CUDA Available: True\n",
"CUDA Version: 13.2\n"
]
}
],
"source": [
"print('CUDA Available:', torch.cuda.is_available())\n",
"print('CUDA Version:', torch.version.cuda)"
]
},
{
"cell_type": "code",
"execution_count": 9,
"id": "390e1b3d",
"metadata": {},
"outputs": [],
"source": [
"research.set_seed(42)"
]
},
{
"cell_type": "code",
"execution_count": 10,
"id": "3849b967",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"polars.config.Config"
]
},
"execution_count": 10,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"pl.Config.set_tbl_width_chars(200)\n",
"pl.Config.set_fmt_str_lengths(100)\n",
"pl.Config.set_tbl_cols(-1)"
]
},
{
"cell_type": "markdown",
"id": "e0f64bf2",
"metadata": {},
"source": [
"# Part 1: ML Model in PyTorch"
]
},
{
"cell_type": "code",
"execution_count": 4,
"id": "f60cc47e",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"annualized_rate=np.float64(93.59487165438073)\n"
]
}
],
"source": [
"# Trading pair symbol\n",
"sym = 'BTCUSDT'\n",
"# time horizon of time series (time interval)\n",
"time_interval = '1h'\n",
"# Max number of auto-regressive lags\n",
"max_lags = 4\n",
"# Forecast horizon in steps \n",
"forecast_horizon = 1\n",
"# Sharpe annualized rate (so it's independent of time frequency)\n",
"annualized_rate = research.sharpe_annualization_factor(time_interval, 365, 24)\n",
"\n",
"print(f\"{annualized_rate=}\")"
]
},
{
"cell_type": "code",
"execution_count": 6,
"id": "a935f32c",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"Downloading BTCUSDT: 100%|██████████| 346/346 [00:11<00:00, 29.86it/s]\n"
]
}
],
"source": [
"start_date = datetime(2024, 10, 29, 0, 0)\n",
"end_date = datetime(2025, 10, 9, 0, 0)\n",
"binance.download_date_range(sym, start_date, end_date)"
]
},
{
"cell_type": "code",
"execution_count": 8,
"id": "282ade73",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"Loading BTCUSDT: 100%|██████████| 346/346 [00:18<00:00, 18.69day/s]\n"
]
},
{
"data": {
"text/html": [
"\n",
"
shape: (8_304, 5)| datetime | open | high | low | close |
|---|
| datetime[μs] | f64 | f64 | f64 | f64 |
| 2024-10-29 00:00:00 | 69939.9 | 69998.4 | 69733.0 | 69939.5 |
| 2024-10-29 01:00:00 | 69939.5 | 70399.4 | 69810.8 | 70287.0 |
| 2024-10-29 02:00:00 | 70286.9 | 71607.0 | 70198.1 | 71173.9 |
| 2024-10-29 03:00:00 | 71173.8 | 71409.0 | 70870.2 | 70927.5 |
| 2024-10-29 04:00:00 | 70927.4 | 71123.4 | 70906.1 | 71062.6 |
| … | … | … | … | … |
| 2025-10-09 19:00:00 | 120600.0 | 121060.1 | 120600.0 | 120897.6 |
| 2025-10-09 20:00:00 | 120897.6 | 121147.2 | 120787.9 | 121045.4 |
| 2025-10-09 21:00:00 | 121045.3 | 121457.2 | 120998.3 | 121407.8 |
| 2025-10-09 22:00:00 | 121407.9 | 121712.4 | 121342.0 | 121619.5 |
| 2025-10-09 23:00:00 | 121619.6 | 121671.1 | 121271.3 | 121579.3 |
\n",
"
shape: (8_304, 2)| datetime | price_median |
|---|
| datetime[μs] | f64 |
| 2024-10-29 00:00:00 | 69848.5 |
| 2024-10-29 01:00:00 | 69969.0 |
| 2024-10-29 02:00:00 | 71180.1 |
| 2024-10-29 03:00:00 | 71170.5 |
| 2024-10-29 04:00:00 | 71029.3 |
| … | … |
| 2025-10-09 19:00:00 | 120839.8 |
| 2025-10-09 20:00:00 | 120942.8 |
| 2025-10-09 21:00:00 | 121300.7 |
| 2025-10-09 22:00:00 | 121500.7 |
| 2025-10-09 23:00:00 | 121448.7 |
"
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"research.plot_static_timeseries(ts, sym, 'close', time_interval)"
]
},
{
"cell_type": "code",
"execution_count": 13,
"id": "ee7b4018",
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"\n",
"\n",
"\n",
""
],
"text/plain": [
"alt.Chart(...)"
]
},
"execution_count": 13,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"alt.data_transformers.enable(\"vegafusion\")\n",
"research.plot_dyn_timeseries(ts, sym, 'close', time_interval)"
]
},
{
"cell_type": "markdown",
"id": "e2b4417c",
"metadata": {},
"source": [
"## Feature Engineering"
]
},
{
"cell_type": "code",
"execution_count": 14,
"id": "fe1ac560",
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"\n",
"\n",
"\n",
""
],
"text/plain": [
"alt.Chart(...)"
]
},
"execution_count": 14,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"price_time_series = pl.DataFrame({'price':[100.0,120.0,100.0]})\n",
"research.plot_column(price_time_series, 'price')"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "497659ef",
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"\n",
"
shape: (3, 5)| price | delta | return | log_return | test |
|---|
| f64 | f64 | f64 | f64 | f64 |
| 100.0 | null | null | null | null |
| 120.0 | 20.0 | 0.2 | 0.182322 | 100.0 |
| 100.0 | -20.0 | -0.166667 | -0.182322 | 120.0 |