Portfolio Construction using Machine Leaning & Optimization theory
Abstract
Over the past decade, data science has played a pivotal role in reshaping the financial industry. This endeavor harnesses historical market data, company fundamentals, recent trends, feature engineering, and a range of pre-trained machine learning algorithms to develop and refine high-yield portfolios. Emphasizing risk reduction to enhance return on investment (ROI), this project employs information retrieval techniques to source fundamental and pricing data from financial platforms. Following data collection, thorough exploratory data analysis is conducted to extract insights from the datasets. Subsequently, feature engineering is implemented, incorporating the Capital Asset Pricing Model (CAPM). The Modern Portfolio Theory (MPT) is then applied to assess asset returns. Utilizing clustering techniques on pricing data, two distinct portfolios are created, which are then merged with company fundamentals to determine overall returns. A variety of regression models, including Linear Regression, Decision Trees, Support Vector Machines (SVM), Lasso Regression, Random Forest, Ridge Regression, Elastic Net, Stochastic Gradient Descent (SGD), and Artificial Neural Networks (ANN), are employed on the merged dataset to evaluate the relative performance of the portfolios.
How to run project
Step# 1: cd portfolioconstructionproj
Step# 2: python manage.py runserver