Skip to content
View Nishu1717's full-sized avatar
1 follower · 0 following

Highlights

  • Pro

Block or report Nishu1717

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don't include any personal information such as legal names or email addresses. Markdown supported. This note will be visible to only you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Nishu1717/README.md

Hey, I'm Nisarg 👋

I'm a data analytics student based in Ontario, currently building projects at the intersection of machine learning, geospatial analysis, and business decision-making. I like working on problems where the output isn't just a model metric but it's a dollar figure, a policy recommendation, or something a non-technical stakeholder can actually act on.

Most of my projects start with a real question: Why does something happen? What is the reason for that? Where does the issue lie? The analysis follows from there.

What I've been building

Finance Risk & Fraud Detection XGBoost classifier on 1M+ credit card transactions. The interesting part wasn't the model — it was building a cost-benefit threshold optimizer that translated predictions into ~$917K in projected savings. Python XGBoost SHAP Scikit-learn

Milton Transit Desert Analysis Spatial analysis of transit coverage gaps in Milton, ON using GTFS stop data and 2021 Census population. Found that 22.9% of residents (~37,670 people) live beyond a 5-minute walk from any transit stop. Python GeoPandas Folium Statistics Canada

Student Churn Early Warning System Random Forest model on 10M+ OULAD clickstream records that flags at-risk students by Week 4 with 70% precision. Behavioral engagement turned out to be a 5x stronger predictor than demographics. Python Scikit-learn Pandas

Loan Process Optimisation Process mining on the BPI 2017 bank loan dataset. Identified "Call incomplete files" as the primary bottleneck and proposed an AI-OCR automation that could cut processing time by 90%. Python Process Mining Pandas Matplotlib

Olist E-Commerce Analytics RFM segmentation and CLV modeling on 100k+ orders from Brazil's largest e-commerce platform. Found that customer satisfaction drops sharply once a delivery is more than 3 days late. Python Pandas Matplotlib Seaborn

Skills

Languages: Python, SQL, R ML & Analysis: XGBoost, Scikit-learn, SHAP, Random Forest, RFM, Process Mining Geospatial: GeoPandas, Folium, GTFS, Shapely Visualization & BI: Streamlit, Power BI, Plotly, Matplotlib, Seaborn Tools: Jupyter, Git, Pandas, NumPy

Get in touch

Popular repositories Loading

  1. Toronto_Smartcity_Insights Toronto_Smartcity_Insights Public

    Interactive dashboard analyzing Toronto's municipal data - crime, 311 requests, and traffic patterns

    HTML

  2. Portfolio Portfolio Public

    Portfolio Website

    HTML

  3. Loan-Process-Optimisation Loan-Process-Optimisation Public

    Data-driven process mining and optimization of a bank's loan funding lifecycle using the BPI 2017 dataset.

    Python

  4. Finance-Risk-Fraud-Detection Finance-Risk-Fraud-Detection Public

    An end-to-end XGBoost fraud detection system optimized for financial cost-benefit analysis ($917k savings) using SHAP for model interpretability.

    Jupyter Notebook

  5. OULAD-Student-Churn-Analysis OULAD-Student-Churn-Analysis Public

    Early Warning System for Student Churn: A machine learning approach using 10M+ OULAD clickstream records to predict at-risk students by Week 4 with 70% precision.

    Jupyter Notebook

  6. olist-ecommerce-analytics olist-ecommerce-analytics Public

    A comprehensive data analytics project on the Olist e-commerce dataset, featuring RFM segmentation, Customer Lifetime Value (CLV) modeling, and logistics performance correlation.

    Jupyter Notebook