|Saturday, February 22|
|PS3 Poster Session III & Continental Breakfast||
Sat, Feb 22, 7:30 AM - 9:00 AM
PPM Recruitment Performance Rate Forecasting (302750)*Renting (Sharon) Xu, Nielsen Audio
Keywords: recruitment performance rate, logistic regression, hierarchical clustering
Arbitron’s Sampling System uses a variety of rates to predict household performance. The recruitment performance rate (RPR) is the percentage of households who agree to participate in a given survey among all households recruited in a certain time period, for a given market and demographic group. Historically, at each planning period, RPR was calculated by accumulating "aggree" households over total households recruited in the past 365 days. The application of a rolling rate does not capture the cyclical nature of performance over time and ignore the impact of demographic characteristics on performance. Therefore, by developing a predictive model with proper attributes from recruited households, Sampling System can forecast better RPR and make smarter selection decisions. In the proposed methodology, the task of forecasting RPR for each market and demographic stratum is accomplished by forecasting an agree/disagree decision for each household. A Logistic Regression model is developed to make a binary choice prediction. Hierarchical clustering method is used to cluster all PPM markets into a few groups to make the model efficient. The model is built and validated by empirical data.