Prophet-Forecasting-For-Portfolio-Optimisation

KO | EN

gitlite — search

#javascript #python #ai #typescript #android #machine-learning #docker #windows #cli #kubernetes #vue #pytorch #security #open-source

Prophet-Forecasting-For-Portfolio-Optimisation

(★ 50)

No description available.

Prophet-Forecasting-For-Portfolio-Optimisation 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# Prophet Forecasting for Portfolio Optimisation

## Project Overview
An end-to-end machine learning project that forecasts stock and asset prices using Facebook/Meta Prophet time series forecasting model, then applies Markowitz portfolio optimisation to rebalance portfolios based on these forecasts.

_(Needless to say it's for illustrative purposes and not financial advice)._

**Live Application**: This project is hosted on a Hostinger VPS, runs every morning at 9am UTC and is accessible at [portfolio-optimisation.com](https://portfolio-optimisation.com)

**Presentation Slides**: [Here](https://gamma.app/docs/Prophet-Forecasting-for-Portfolio-Optimisation-7qsgynwy1h5x3it) are slides to accompany this project.

## Components

### 1. Prophet (Time Series Forecasting)

**What is Prophet?**

Prophet is Facebook's open-source time series forecasting tool designed for business forecasting. It handles trends, seasonality, and holidays automatically, making it robust and easy to use for forecasting time series data.

**How It Works in This Project:**

- Input: Historical price time series with datetime index
- Model: Prophet fits additive components (trend, seasonality, holidays)
- Output: Forecasted prices for each asset in the portfolio for the next trading day
- Training: The model fits to historical price data and generates one-step-ahead forecasts

### 2. Markowitz Portfolio Optimisation

**What is Markowitz Portfolio Optimisation?**

Markowitz portfolio optimisation, also known as Modern Portfolio Theory (MPT), is a mathematical framework for constructing optimal portfolios. Developed by Harry Markowitz in 1952, it balances the trade-off between expected returns and risk.

**Key Concepts:**

- **Expected Return**: The weighted average of expected returns of individual assets
- **Risk (Volatility)**: Measured as the standard deviation of portfolio returns
- **Correlation**: How assets move relative to each other
- **Efficient Frontier**: The set of optimal portfolios offering the highest expected return for a given level of risk

**The Optimisation Problem:**

```
Maximize: μᵀw - λ(wᵀΣw)

Subject to:
- Σwᵢ = 1 (weights sum to 1)
- wᵢ ≥ 0 (long-only portfolio, optional)
- Additional constraints (sector limits, etc.)
```

Where:
- `μ` = vector of expected returns (from Prophet price forecasts)
- `Σ` = covariance matrix of asset returns
- `w` = portfolio weights
- `λ` = risk aversion parameter (configurable in `src/settings.py`)

**How It Works in This Project:**

1. **Input**: Forecasted returns (derived from Prophet price predictions) for each asset
2. **Risk Estimation**: Historical covariance matrix calculated from asset returns
3. **Optimisation**: Solves for optimal weights that maximise risk-adjusted returns using SciPy's SLSQP solver
4. **Output**: Recommended portfolio allocation (weights for each asset)
5. **Rebalancing**: Portfolio is rebalanced based on these optimal weights

## Project Workflow

```
Historical Data Extraction
    ↓
Data Preprocessing
    ↓
Prophet Model Training
    ↓
Price Forecasting
    ↓
Markowitz Optimisation
    ↓
Optimal Portfolio Weights
    ↓
Results Saved to Supabase
    ↓
Streamlit Dashboard Hosted on Hostinger VPS
```
## Installation

### Standard Installation

```bash
# Install dependencies using Poetry
make install-dev

# Or manually
poetry install
```

### Requirements

- Python 3.12+
  - I recommend installing through [PyEnv](https://github.com/pyenv/pyenv)
  - PyEnv can be installed through [Brew](https://brew.sh/).
- Poetry
  - [Basic usage](https://python-poetry.org/docs/basic-usage/)
- CircleCI account
  - [Setup guide](https://circleci.com/blog/setting-up-continuous-integration-with-github/)
- Supabase account and project
  - [Starting guide](https://supabase.com/docs/guides/getting-started)
- [Hostinger VPS](https://www.hostinger.com/vps-hosting)
  - [Guide to deploying a Streamlit App on Hostinger VPS](https://egorhowell.notion.site/Streamlit-Deployment-Guide-on-Hostinger-VPS-2ad2dbb15bea808c9683f6da61e3a4e8?source=copy_link)

## Usage

### Basic Usage

```bash
poetry run python -m src.main
```

Or using the Makefile:

```bash
make run
```

### Configuration

Edit `src/settings.py` to customise:

- **Portfolio Tickers**: Modify `PORTFOLIO_TICKERS` list
- **Risk Aversion**: Adjust `RISK_AVERSION` (higher = more risk averse)
- **Minimum Allocation**: Change `MINIMUM_ALLOCATION` (minimum weight per asset)
- **Date Range**: Update `START_DATE` and `END_DATE` for historical data

Example:

```python
# src/settings.py
PORTFOLIO_TICKERS = ["AAPL", "MSFT", "GOOGL", "TSLA", "AMZN"]
RISK_AVERSION = 3 
MINIMUM_ALLOCATION = 0.05 
START_DATE = "2024-01-01"
```

### Programmatic Usage

```python
from src.main import run_optimisation

result = run_optimisation(
    tickers=["AAPL", "MSFT", "GOOGL"],
    start_date="2024-01-01",
    end_date="2024-12-31"
)

print(f"Optimal Weights: {result['weights']}")
print(f"Predicted Returns: {result['predicted_returns']}")
print(f"Current Prices: {result['current_prices']}")
print(f"Prediction Date: {result['prediction_date']}")
```

### Running the Streamlit Dashboard

The Streamlit dashboard reads from Supabase to display historical predictions, portfolio weights, and performance metrics:

```bash
poetry run streamlit run src/streamlit_app.py
```

Or using the Makefile:

```bash
make dashboard
```

The dashboard allows you to:
- View portfolio weights and predictions for any date
- Analywe individual stock performance over time
- Compare predicted vs actual prices
- Track prediction accuracy metrics

**Note:** The dashboard requires Supabase to be configured and populated with data from previous optimization runs.

Prophet-Forecasting-For-Portfolio-Optimisation

Prophet-Forecasting-For-Portfolio-Optimisation 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Prophet-Forecasting-For-Portfolio-Optimisation

Prophet-Forecasting-For-Portfolio-Optimisation 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README