Industrial Projects

2008-09 Projects

Multiple Target Tracking

Sponsor: Lincoln Laboratory
Advisors: Jon Abraham, Ted Clancy (ECE)
Students: Matthew Connor, Kathleen Haas, Alexander Volfson

Due to radar's range measurement accuracy, Range Time plots are used to represent radar data. When objects' tracks cross on a Range Time plot, it is uncertain which track belongs to which target. An analysis of the frequency and angle of these crossings was performed. Mathematical analysis concluded that in certain situations, only one type of crossing can result. Further Monte Carlo simulations were used to study these crossing statistics in other situations.
In addition, it was examined how well targets could be tracked through an individual crossing. The probabilities of correct track association were calculated as a function of a variety of factors. Given our models and assumptions, sensor fusion of Range Time and Range Doppler analysis substantially improved crossing classification.

Demographic Assumptions

Sponsor: Hewitt Associates
Advisor: Jon Abraham
Students: Theresa Cheng, Ben Gilde, Charlotte Paige McAuliffe

Hewitt Associates, in order to perform valuations on pension plans, must make certain demographic assumptions. The objective of this project was to choose statistical tests to analyze the accuracy of these assumptions, and to provide the sponsor with an automated tool to use on an ongoing basis. This MQP created a tool within Microsoft Excel that works with Hewitt's current pension valuation software to analyze the statistical accuracy of Hewitt's demographic assumptions. The analysis compares how well Hewitt's expected demographic probabilities performed using exact confidence intervals and the calculation of p-values adjusted for False Discovery Rate. This tool flags assumptions that need to be reassessed and will aid Hewitt in improving the accuracy of their demographic assumption.

Analyzing Trends in Loss Frequency and Severity

Sponsor: The Hanover Insurance Group
Advisor: Jon Abraham
Students: Ethan Brown, Norman Lam

Hanover Insurance evaluates historical data to analyze trends in loss frequency and severity of claims. The trends are caused by external factors, such as legislative, environmental and economic forces. Trends were analyzed using two different approaches, one correlating the trends from prior data to external factors, and another comparing the impact of events to trends in the data. The analysis mathematically quantified the effect of each external force and isolated factors which were most significant to the trends.

Methods of Forecasting Future Losses

Sponsor: The Hanover Insurance Group
Advisor: Jon Abraham
Students: Haiying Liu, Aryeh Shatz, Mingzhu (Julia) Zheng

This project finds the best mathematical model to forecast the future losses for each state from a given series of historical data by Hanover Insurance Group. There are nine models chosen after analyzing all the states' data. The score method is applied to seek out the accurate models among the nine models. Comprehensive analysis also shows that some external experience needs to be used in order to make the prediction more precise.

Statistical and Experimental Analysis of a Torque Model for Self-tapping Screws

Sponsor: Bose Incorporated
Advisor: Suzanne Weekes
Student: Kate Boulanger

A torque study is a series of lab experiments to determine the force needed to insert a screw into a given material. It is important to determine the torque needed to form threads and seat the screw (drive torque). Past MQPs have produced mathematical models of the self-tapping screw insertion process and developed a graphical user interface that takes in screw and material properties and produces a torque curve. In this work we performed 8 torque studies. We compare lab drive torques with WPI model drive torques using various statistical tests. We consider the effect of using a filter to smooth out the lab data. We investigate the modes of failure. We also determine the sensitivity of the WPI model. On this joint we find that the WPI drive torques fit well with the lab drive torques.

Statistical Multi-Source Predictive Models and Error Estimates: Major USDA Crop Protection Forecasts and Estimates

Sponsors: National Institute for Statistical Sciences and National Agricultural Statistics Service
Student: Criselda Toto

My research last summer is part of the Cross-Sector Research in Residence Program established by the National Institute of Statistical Sciences (NISS) in partnership with the National Agricultural Statistics Service (NASS), the survey and estimation arm of the U.S. Department of Agriculture. This new collaborative venture by NISS and the USDA is the first project of a NISS initiative to host academic-government research teams focused on specific federal agency objectives. Three teams, each composed of five people (a faculty researcher in statistics, a NASS researcher, a NISS mentor, a postdoctoral fellow and a graduate student) were formed to work intensively together at NISS during the summers of 2009 and 2010 to solve research questions posed by NASS. The research problem assigned to my team is entitled: "Statistical Multi-Source Predictive Models and Error Estimates: Major USDA Crop Protection Forecasts and Estimates."
The main goal of our team's research project is to improve the process of producing multiple forecasts of crop protection throughout the growing season and estimates production at the end-of-season or after harvest by building a statistical model using information collected from multiple sources (USDA surveys and administrative/auxiliary information, including weather and remotely sensed data). At present, these information are synthesized by a panel of experts in USDA’s Agricultural Statistics Board (ASB) to come up with the official forecasts/estimates that are published. Our aim is to obtain a more objective process of obtaining official forecasts using data modeling.

Maintained by
Last modified: Jun 20, 2010, 09:03 EDT
[WPI] [Math] [Home]