fnine99/markowitz_mpt

Markowitz efficient frontier model in modern portfolio theory

This model is made for learning puposes only and should not be used for investment or decision making.

Portfolio risk management

An application of the Markowitz efficient frontier model and the Modern Portfolio Theory

In this project I implement the almost complete automation of the portfolio optimization and risk using the Markowitz model.

Requirements:

python3 and pip correctly installed

Usage:

git clone https://github.com/Fnine99/Markowitz_mpt

cd Markowitz_mpt

pip install -r requirements.txt

python3 main.py

Content description:

Step 1) 5 years monthly price time series data fetching on the Twelve Data API

see data.py

Step 2) Various methods on each assets including:
-Monthly prices
-Monthly returns
-Arithmetic mean return
-Geometric mean return
-Monthly returns standard deviation

see assets.py

Step 3) Portfolio construction and Various portfolio methods including:
-Portfolio return
-Porfolio covariance_matrix
-Portfolio variance
-Portfolio standard deviation
-Portfolio correlation matrix
-Portfolio inverse covariance matrix

see portfolio.py

Step 4) Portfolio optimization with Scipy algorithms including finding the assets weights which:
-Minimize the portfolio return
-Maximize the portfolio return
-Minimize the portfolio variance
-Maximize the portfolio variance
-Maximize the portfolio Sharpe ratio

see optimize.py

Step 5) Efficient frontier construction and portfolios modelling including:
-Generate X number(1M) of portfolios
-From those generated portfolios locate the assets weights which:
>Minimize the portfolio variance
>Maximize the portfolio Sharpe ratio
-Plot step 4 and 5

see frontier.py

Results Example

Step 2:

Monthly Prices:

Monthly returns:

Step 3:

Portfolio Covariance Matrix:

Portfolio Correlation Matrix:

Portfolio Inverse Covariance Matrix:

Step 4:
Note that the portfolio with asset weight bounds of [0.000, 0.900] and with a risk-free rate of 0.045. Very interesting to see that, when generating 1M of portfolios, we can very precisely predict the optimized portfolios.






Step 5: