|
One of the core pieces of the industrial mathematics curriculum is the portfolio of projects accumulated over a two-year period. Below, you can find the projects that I have worked on. In each case, a PDF of the project results is available by clicking on the PDF icon. Please e-mail me if you have any questions or comments regarding these projects. In the future, I would like to use blog posts to explain some of the mathematics behind these projects; in particular, the Digital Watermarking and the Digit Recognition papers.
Abstract: ACO Hardware has gathered a large amount of diverse and conflicting financial data combined with site, demographic, traffic, consumer behavior and competition data in an effort to build a "successful store profile." This profile will then be used as a tool for evaluation when developing a new location and expanding or moving an underperforming location. The focus of this project is to evaluate ACO Hardware's current operations, extract the relevant data, and develop a ten-factor model to evaluate success.
Abstract: ACO Hardware has contracted with ADVO to obtain detailed consumer spending profiles from a variety of census and economic data by zip code. Each profile is characterized by average household size, age breakdowns, economic conditions, education levels, and marital status. The focus of this project is to compare the gathered data with actual spending habits and dollars spent to determine a "successful customer profile" using a least-squares regression. The goal is to have a predictive model that uses the best demographic mix to determine profitable new store locations.
Abstract: With the explosion of digital age has come an explosion in illicit copying. Numerous protection schemes have been developed to combat such copying. One of these techniques is known as digital watermarking, where a "watermark," personally identifiable information, is hidden within a copyrighted work. This paper concerns one technique for hiding such information, the singular value decomposition.
Abstract: Handwritten digit recognition is an important and challenging problem in pattern recognition. This paper reports on experiments done on the MNIST set of handwritten digits, using two different feature spaces and a variety of classifiers on each space. Performance is compared to benchmarks in the field.
Abstract: The ability to reasonably forecast price movements within the chemical market is of great importance to many departments within the Dow Chemical Company. This project is the first phase of two in the development of a pricing model to perform such forecasting. The focus of this phase is to apply the agent-based modeling paradigm to simulate a theoretical commodity market found in John D. Sterman's "Business Dynamics: Systems Thinking and Modeling for a Complex World." Sterman provides a price model based on the system dynamics paradigm that assumes a group of companies composing a market all behave alike and thus can be modeled as a single unit. The agent-based model is written in Java and based on the framework provided by the Repast toolkit.
|
