Evaluating Yahtzee Strategies Using Simulation

This is a solo project completed for the course ISYE 6664 (Simulation and Modeling for Engineering and Science) for my master’s at Georgia Tech. This paper details how I simulated the game Yahtzee in Python to evaluate three distinct playing strategies: random selection, prioritizing the highest score per turn, and prioritizing upper section scoring. By comparing the score distributions generated by these strategies through data visualization and statistical tests, my study concluded that both prioritizing the highest score and prioritizing the upper section result in significantly higher scores than random selection, highlighting the advantage of strategic play in Yahtzee.

Impact of Zone and Construction Permits on Real Estate Prices in Washington, D.C.

This is a group project completed for the course MGT 6203 (Data Analytics in Business) as part of my graduate studies at Georgia Tech. The objective of this project was to evaluate how neighborhood investment and development, as measured by construction and building permits, impact residential sale prices within Washington, D.C. over a five year period from 2018 – 2022. Data from publicly accessible sets released by the Washington, D.C. government were utilized to scrutinize patterns at the Census Block Group level.

Holt-Winters Forecasting with CVG Temperature Data

The goal of this exercise is to use Holt-Winters forecasting, also known as triple exponential smoothing, to predict future monthly average temperatures in the Cincinnati area. The data was taken from the National Weather Service, and contains the monthly temperature averages in Fahrenheit from January 2000 through December 2022.
⁣

I used the great e-textbook Forecasting: Principles and Practice by Hyndman and Athanasopoulos to inform my knowledge and methodology. The link below is my initial analysis – further below in this section is my retrospective comparison to the actual 2023 monthly temperatures.

Revisiting this project after the initial forecast, I’ve plotted my forecasted values against what was actually recorded for 2023. Overall, it’s fairly close, with the exception of January, February, and December. However, I believe this is not an issue with the model, as NOAA did note that 2023 was hotter than normal. January was the 6th hottest on record and February was the 4th hottest. December was also noted as the warmest on record.

Titanic Classification with Support Vector Machines

This is the (classic) Titanic classification problem. The training data set contains information about each passenger, and whether they survived the sinking of the Titanic or not. Let’s see how accurate of a model I can make, using Support Vector Machines classification models with K-fold Cross Validation, to predict if a passenger would survive or not.
⁣

The training data is provided by Kaggle.com, from their Titanic – Machine Learning from Disaster competition. There was testing data provided, but due to the nature of the competition, it does not provide the survival column. Because of this, the testing data will be created from a split in the training data.

Import Container Tracker

This is a script I designed to make it much easier to track shipping containers in bulk. It takes an input .txt file of desired container numbers and drives a Chrome window to input the numbers into SeaRates.com and return the last status in a .csv file.

If the site is slow or the container number does not yield any results, the script will wait 20 seconds before yielding “No useful information can be output” and moving on. In these instances, I have to manually enter the container number into a separate site (typically that of the actual freight forwarder).