Mathematical Decision Making: Predictive Models and Optimization

Survey the extraordinary range of applications for operations research and predictive analytics. Professor Stevens defines these fields, previews the mathematical techniques that underlie them, and charts their history, from World War II defense research to their rapid growth in the computer era.

Linear regression is a powerful method for describing connections between related quantities. Analyze several problems using linear regression. For example, predict the waiting time for an eruption of the Old Faithful geyser based on how long the previous eruption lasted.

Explore more complex linear regression problems, which involve nonlinear functions and/or multiple inputs. Many real-life situations require these approaches, called transformation of variables and multiple linear regression. Learn how to envision the data graphically, and witness the ease with which spreadsheets solve these problems.

Time series forecasting is a valuable tool when there's little data on what drives a process. Using the example of U.S. housing starts, learn how to glean information from historical figures, taking into account both long-term trends and seasonal fluctuations to create a forecast and assess its reliability.

Plunge into the fast-growing field of data mining, which exploits computational power and innovative algorithms to analyze the ever-increasing deluge of data. Focus on classification and prediction, seeing how classification trees can help solve the problem of building a filter that predicts spam email messages.

Delve deeper into data mining by exploring affinity analysis, or "what goes with what." One approach uses association rules to discover relevant connections between variables, while another employs clustering. For example, Pandora Radio uses these tools to make music recommendations based on a listener's song preferences.