Shaneeza Hasnani

FEATURED · Event Study

Python Pandas Finance

Fraud Disclosure Event Study

I wanted to know how markets actually respond when corporate fraud goes public. Using the Fama-French Three-Factor Model, I measured abnormal returns around disclosure dates and found patterns most people overlook.

◈ Fama-French r² = 0.87 · CAR significant at p<0.01

View Details GitHub ↗

Fraud Detection

Snowflake SQL Python

Fraud Detection in Snowflake

Designed a Snowflake-native fraud detection workflow that engineers transaction features, scores risk, and surfaces high-risk activity for review without moving data out of the warehouse.

◈ Snowflake feature engineering · Risk scoring pipeline

View Details GitHub ↗

Machine Learning

Python Scikit-learn

Titanic Survival Prediction

A classic dataset, but my focus was on doing it right. Built a full preprocessing pipeline, engineered features from raw passenger data, and evaluated with accuracy, AUC, and F1 to avoid misleading metrics.

◈ Accuracy: 82% · AUC: 0.89 · F1: 0.81

View Details GitHub ↗

Data Analysis

R ggplot2 EDA

WMATA Ridership Analysis

Pulled 5 years of DC metro data and broke down ridership patterns by station, season, and whether it was a holiday. Found things that genuinely surprised me about how commuter behavior shifts.

◈ 5-year dataset · R² = 0.93 · Seasonal decomposition

View Details GitHub ↗

Statistical Analysis

R Regression

NYC Job Salary & Career Analysis

Scraped and modeled NYC job posting salaries to understand what career level actually explains. Multiple regression with hypothesis testing on 2,400+ postings. The results were less obvious than you'd expect.

◈ Adj. R² = 0.71 · n=2,400 jobs · p<0.001

View Details GitHub ↗

Labor Analytics

R Data Mining

LinkedIn Skills & Salary Analytics

Used k-means clustering on LinkedIn job data to find which skill combinations actually drive salary. A personal project that started as curiosity about my own career path.

◈ k=5 clusters · Silhouette = 0.68

View Details GitHub ↗

Regression Analysis

R Modeling

Spotify Audio Features Analysis

Regression analysis on 5,000 Spotify tracks to see how much audio features like tempo, valence, and genre actually explain energy levels. Turns out acousticness is more predictive than genre.

◈ RMSE = 0.09 · 5k tracks · R² = 0.76

View Details GitHub ↗