Portfolio Optimization on NIFTY Thematic Sector Stocks Using an LSTM Model

TL;DR

This paper develops an LSTM-based model to predict stock prices and designs optimal portfolios for Indian sector stocks, demonstrating high prediction accuracy and effective portfolio optimization over a five-year period.

Contribution

It introduces a novel approach combining LSTM predictions with portfolio optimization for Indian sector stocks, which is less explored in existing literature.

Findings

LSTM model achieved high prediction accuracy for stock prices.

Optimized portfolios showed favorable risk-return profiles.

Predicted returns closely matched actual returns after seven months.

Abstract

Portfolio optimization has been a broad and intense area of interest for quantitative and statistical finance researchers and financial analysts. It is a challenging task to design a portfolio of stocks to arrive at the optimized values of the return and risk. This paper presents an algorithmic approach for designing optimum risk and eigen portfolios for five thematic sectors of the NSE of India. The prices of the stocks are extracted from the web from Jan 1, 2016, to Dec 31, 2020. Optimum risk and eigen portfolios for each sector are designed based on ten critical stocks from the sector. An LSTM model is designed for predicting future stock prices. Seven months after the portfolios were formed, on Aug 3, 2021, the actual returns of the portfolios are compared with the LSTM-predicted returns. The predicted and the actual returns indicate a very high-level accuracy of the LSTM model.