Data Science and Analytics

Project 1: Portfolio Optimization

Implemented a algorithm to find the optimal allocation of assets that aims to maximize returns while minimizing risk.

Curated 10 low-correlated stocks across various sectors to enhance portfolio diversity and reduce risk.
Assessed the percentage of free floated stocks to ensure adequate liquidity and market representation.
Conducted Monte Carlo simulations on 5 year data and calculated Sharpe ratio to identify optimal portfolio configurations.
Visualized risk-return tradeoff profile through efficient frontier, revealing a best Sharpe ratio of 1.62 and 17.8% return.

Build an in-house risk model on behalf of a bank to make lending decisions for subprime mortgages to maximize profitability.

Implemented a Random Forest model to predict borrower’s probability of default (PD) attaining an AUC score of 0.95.
Designed a credit scorecard that assigns scores based on 29 key metrics, effectively evaluating customer creditworthiness.
Classified credit scores into 10 distinct levels, enabling easy interpretation and decision-making for lending officers.
Computed cumulative loss and coverage metrics, determining a threshold score of 550 to maximize profitability.

Conducted Data Analysis on financial complaints dataset to identify the areas for improvement and aims to improve customer experience.

Study a dataset containing 6 lakh+ complaints received by various finance companies over 5 Years (from Dec-2011 to Nov-2016).
Analyzed different attributes and compared last 5 years complaints to last 1 year to evaluate variation in previous years.
Created an Interactive Dashboard using Tableau to enhance stakeholders’ comprehension of the problem.

Performed customer classification to subscribe for the term deposit for Targeted Marketing and Improved Conversion Rates.

Implemented Feature Selection using Extra Tree Classifier and check p-value statistic test for every feature.
Converted imbalanced dataset (9 : 1) into balanced dataset (1 : 1) by executing oversampling using SMOTE.

Conducted Data Analysis on IPL matches to derive actionable insights and trends, contributing to strategic decision-making

Performed Exploratory Data Analysis on Indian Premier League dataset containing 1.8 lakh records of each delivery.
Identified out Most Valuable Players and Teams a company should endorse for their products to increase their sales.
Identified key factors like toss results and fielding potential should be considered as they had impacts on match results

Developed content based recommendation system that provides movie suggestions based on user preferences.

Merged two datasets of 4800 movies and dug out suitable features for our system and performed Data Preprocessing.
Combined all important features (genres/keywords/actors/directors) into one column and converted them into vectors.
Model was trained to find top-5 similar movies on the basis of distances between each vectors using Cosine similarity.