Data Collection & Preparation The total cost for the entire tower dataset was only $49.
USBizData was the only purchased data source; it included addresses and approximate number of employees for most businesses in Wisconsin. Data collection & preparation accounted for at least 70% of the overall project time.
Predictive and Explanatory Linear Regression Models PREDICTIVE EXPLANATORY R-Squared: 54% 40% PRESS 46% 34% Independent Variables: - Population (County) - Business Indicator - Lake Indicator - Population (County) - Business Indicator - Lake Indicator
Boosted Neural Network
Out of sample range: 35% to 73% with an average value of 57%
- Percent of Caucasians (County) - Cable Internet Above 100 mbps - Count of Neighboring BTW Towers - Height
- Percent of Caucasians (County) - Cable Internet Above 100 mbps - Count of Neighboring BTW Towers
- Percent Homeowners (Tract) - Percent Population Over 65 - Percent of People with at least a Bachelor’s Degree (County) - Percent of People with at least a High School Diploma (County) - Mean Travel time to Work (County) - Neighboring Business Count - Satellite Internet Speed Above 100 mbps
Variable Importance
To better understand the importance of each variable, relative importance was computed for the 6 predictors in the explanatory model.
Central Wisconsin Report - Fall 2021
21
Made with FlippingBook Learn more on our blog