FERM Project

Topic - Machine learning for portfolio diversification and allocation

Group members

1. Aayushmaan Jain (J022)
2. Pratyush Patro (J047)
3. Devanshu Ramaiya (J050)
4. Ishani Shah (J067)
5. Amit Prajapati (J075)

Part 1: Machine Learning for portfolio diversificaion

This part uses K-Means clustering to diversify the portfolio into three different clusters

1. Low Risk High Return
2. High Risk High Return
3. High Risk Low Return

Part 2: Monte Carlo Simulation for portfolio allocation

This part uses monte carlo simulation to allocate weights to the top 20 stocks (according to sharpe ratio)
For the purpose of computational efficiency, the simulation is done via matrix multiplicaion

Data - $(n_{days}, n_{stocks})$
Weights - $(n_{stocks}, n_{simulations})$
Simuation matrix - $(n_{days}, n_{simulations})$

The simulation matrix contains the daily data for every simulation column wise which can then be aggregated to calculate the returns and volatility

Demo on Google Colab