Enterprise BI solution
Created an end-to-end BI system for a product manufacturing company, which can be used to analyze its historic sales data and derive actionable insights using performance metrics like revenue, costs, profit, target profit margins.
Email spam detection and classification
Conducted a comparative study to evaluate the performance metrics of 3 email spam classifiers implemented using naïve bayes, decision trees, and random forest Machine Learning algorithms and trained on Gmail mail text data.
United Airlines incident impact analysis
Conducted an exploratory impact analysis of the United flight 3411 incident and understand if there was any correlation between the user tweets and NYT articles sentiment about the incident and UAL stock prices.
Neural net for digit classification
Implemented a 2-layer neural network using gradient descent to classify handwritten digits (0-9) with 85% test accuracy. Used the MNIST handwritten digits dataset for model training and testing. Used tanh and sigmoid as the activation functions.
Movie and Actor popularity analysis
Analyzed a corpus movie and actor data scraped using the open movie DB API to generate the popularity trend of various actors over time. The metric i used to measure popularity of the actor was the cumulative profit that the actors films made in a particular year. I also analyzed the popularity of the various movie genres to understand the seasonal popularity of different movie genres. Used the Bokeh library to create the visualizations.
Broadband speeds in Seattle Analysis
Visualization created using the D3 JavaScript library which helps compare the costs of different broadband providers in Seattle for different upload and download speed bands.
Diamond prices regression analysis
Created a regression model to predict diamond prices based on variation in predictor values like carat, cut, color, clarity, and dimensions of the diamond. Tools used – R, Shiny.
Online adaptive exam system
Created an online exam system using a personalized learning algorithm based on exponential smoothing which adapts the difficulty based on student performance. Implemented it using PHP, JavaScript, and MySQL.