📊

My ML Journey

2022 — First steps into Machine Learning & Data Science

🌐 Public Repository — View on GitHub →

This is where my journey into AI and data science began. A collection of projects working through classic ML problems — from exploratory data analysis to building neural networks from scratch.

📑 Projects

🚨 911 Calls Analysis (13 plots) 📢 Advertising (6 plots) 🏠 Housing Prices (5 plots) 🛒 E-commerce (7 plots) 📈 Bank Stocks 🔢 MNIST (2 plots) 🚢 Titanic (22 plots) 🧠 Deep Learning Coursera
🚨

911 Calls Exploratory Data Analysis

~100,000 emergency calls from Montgomery County, PA

Key Findings

  • • EMS: 48,877 calls (49%) — Most common
  • • Traffic: 35,695 calls (36%)
  • • Fire: 14,920 calls (15%)
  • • Lower Merion township has the most calls

All Visualizations

Calls by Reason Day of Week Month Monthly Trend Linear Fit Daily Calls Traffic Fire EMS Heatmap Hour Clustermap Hour Heatmap Month Clustermap Month
Pandas Seaborn EDA
📢

Advertising Click Prediction

Logistic Regression for ad click prediction

Predicting whether users will click on ads based on time on site, age, income, and internet usage. Clear separation visible between clickers and non-clickers.

All Visualizations

Age Histogram Distribution Income vs Age Time vs Age Time vs Usage Pairplot
Logistic Regression Scikit-learn
🏠

USA Housing Price Prediction

Linear Regression for price estimation

💡 Key Insight

Area income is the strongest predictor of house price. Model achieves good fit with normally distributed residuals.

All Visualizations

Distribution 1 Distribution 2 Heatmap Predictions Residuals
Linear Regression Scikit-learn
🛒

E-commerce Customer Analysis

Should they focus on mobile app or website?

💼 Business Recommendation

Mobile app drives more revenue per minute. Either develop the website to catch up, or double down on the app.

All Visualizations

Website vs Spending App vs Spending App vs Membership Pairplot Linear Fit Predictions Residuals
Linear Regression Business Analytics
📈

Bank Stock Analysis

Analyzing the 2008 Financial Crisis

Analysis of Bank of America, CitiGroup, Goldman Sachs, JPMorgan, Morgan Stanley, and Wells Fargo during and after the 2008 crisis.

Analysis

  • • Daily returns & risk assessment
  • • Correlation heatmaps
  • • Moving averages & Bollinger Bands
  • • Candlestick charts

Findings

  • • 2008 crash visible across all banks
  • • CitiGroup most volatile
  • • High correlation between banks
pandas_datareader Cufflinks Technical Analysis
🔢

MNIST Digit Classification

The "Hello World" of Deep Learning

My first neural network! Classifying 70,000 handwritten digits (28×28 grayscale) into 10 classes.

Visualizations

Sample Digits Results
Neural Networks TensorFlow
🚢

Titanic Survival Prediction

The classic Kaggle challenge

💡 Key Insights

  • • "Women and children first" — females had much higher survival
  • • Class matters — 1st class survived more
  • • Higher fare = better survival

All Visualizations (22 plots)

Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Plot 7 Plot 8 Plot 9 Plot 10 Plot 11 Plot 12 Plot 13 Plot 14 Plot 15 Plot 16 Plot 17 Plot 18 Plot 19 Plot 20 Plot 21 Plot 22
Logistic Regression Feature Engineering Kaggle
🧠

Deep Learning Specialization

Andrew Ng's Coursera Course

Building neural networks from scratch — no frameworks, just NumPy.

Planar Data

2-layer network for non-linear classification

DNN from Scratch

Forward prop, backprop, gradient descent

Cat Classifier

L-layer network for image classification

NumPy Backpropagation From Scratch

🎓 What I Learned

  • • Data manipulation with Pandas
  • • Visualization with Matplotlib & Seaborn
  • • ML models with Scikit-learn
  • • Neural networks from scratch
  • • EDA best practices
  • • Feature engineering
  • • Model evaluation metrics
  • • Business insights from data
← Back to Journey View on GitHub →