The Schedule Effect: can recurrent peak infections be reduced without vaccines, quarantines or school closings? (2013)
by Danilo Diedrichs, Paul Isihara and Doeke Buursma
Using a standard SIR (Susceptible-Infected-Recovered) model with seasonal dynamics, we study the "schedule effect", which allows for a significant reduction in recurrent peak infections of endemic diseases in schools by varying the traditional school calendar. Analysis of the phase plane explains the relationship between the maximum recurring infection peaks and the period of an oscillating transmission function. The response may exhibit period-doubling and chaos induced at certain periods, leading to increased peaks. We show how to take these effects into consideration to design an optimum school schedule.
A Mathematical Model for the Growth and Decline of the Church in DuPage County (2013)
by Daniela Cuba
The project analyzes the population dynamics that affect size of the church in DuPage County, Illinois. A dynamic model (which we name the UEB model), similar to the SIR model used in epidemiology, divides the population into compartments of Unbelievers, Enthusiasts (who actively bring unbelievers into the church), and passive Believers. In addition to the conversion dynamics, our model also incorporates the long-term demographic fluctuations of DuPage County. The parameters of the model are determined by fitting church and demographic data obtained from the Census Bureau. The results of this study are useful to identify the most impactful strategies for churches to increase their membership.
Constrained Optimization Model for Quantitative Criminology (2013)
by Korey Clement
The constrained optimization model for quantitative criminology, first introduced by criminologists Alfred Blumstein and Daniel Nagin, is used to control and minimize the crime rate of a given population. Using the most recent data on crime and punishment available (2009), we use this model to determine the lowest crime rate that can achieved in the United States. The sensitivity analysis of the model's parameters reveals what steps must be carried forth in order to reduce the crime rate to a global minimum.
Development of an Application for Indoor Temperature Control Efficiency (2013)
by Roland Hesse
Using numerical techniques to discretize and solve heat equation (a partial differential equation) in three dimensions, we devise a computerized application that models the heat flow and determines temperature gradients in a building. The application allows for the geometry of the rooms and insulation properties of the boundaries to be specified, as well as the indoor locations where people are most likely to be found. We use this system to locate the optimal placements of HVAC (Heating, Ventilation, and Air Conditioning) for overall efficiency in temperature control and reduction of wasted energy and climate-control costs.
Computational Composition of Traditional Scottish Music (2013)
by Tim Macdonald
Traditional Scottish music has recurring patterns at the harmonic, melodic, and structural levels. Using a combination of automated pattern recognition techniques and domain-specific knowledge, we develop a system that, seeded with a corpus of existing tunes, composes original music in the same style. This is accomplished using a long short term memory network---a type of recurrent neural network. The corpus used was the complete works of 18th century composer William Marshall, which was transformed into a sequence of integers suitable for inputting into the network. Post-conversion, the music was used to train the network, and the trained network was used for generating new music.
Inventory Models in Disaster Relief (2012)
by Nate Veldt
Models for supply chain management can be used to assist humanitarian relief organizations in calculating how to efficiently provide for a population affected by a natural disaster. This project explores in detail a single period probabilistic model for fulfilling a demand while minimizing costs. The model is implemented in MATLAB and several examples are given of how this model might be used in a specific disaster relief situation. A Monte Carlo simulation is used to generate potential values for demand to then analyze how this model might be used to meet a demand that stretches over multiple periods.