ritvikgupta199 / Portfolio-Optimisation-using-Clustering

Portfolio Optimisation using Clustering via Ball Mapper Algorithm and MPT

Two stage Portfolio Optimisation using Ball Mapper Algorithm

Report

For Clustering

cd BallMapper
g++ ball.cpp -o ball
./ball <year> <qtr> <eps>

where <year> and <qtr> are the quarter for which clustering is to be done and <eps> is the epsilon parameter for the Ball Mapper Algorithm.

For Portfolio Optimisation

cd MeanVariance
python3 max_sharpe.py --year <year> --quarter <qtr> --select_num <select_num>
python3 meanvar.py --year <year> --quarter <qtr> --model <model>
python3 eval_portfolio.py --year <year> --quarter <qtr> --dir <dir> --model <model>

Here,

• <year> and <qtr> are the quarter using which the clusters have been created
• <select_num> is the number of stocks to be picked from each cluster based on Sharpe ratio
• <model> is "max_sharpe" or "min_vol" based on the optimisation objective
• <dir> is the location where the evaluation graphs are saved

For evaluating the method for each of the avaliable quarters, run run.sh.

Data sources

• Company Financials and stock prices from AlphaVantage API
• S&P 500 Index values from FRED