The turf is always greener: Predicting decommitments in college football recruiting using Twitter data

Highlights：

• Incorporating Twitter network data adds value to American college football decommitment predictions.

• New connections to other schools are associated with significantly increased odds of decommitment.

• Out-links to future teammates are strong predictors of decommitment.

• In-links from coaches and fellow recruits are strong predictors of decommitment.

• Social media data may be useful for predicting turnover in other domains.

摘要

We utilize the wealth of data related to American college football recruitment as a laboratory for studying the impact of social networks on organizational turnover. We combine data about athletes' recruiting activities and college choices with data from their Twitter social networks to predict decommitments over time - specifically, which athletes will decommit from their current college in a given month. Our results demonstrate the value of considering online social networks for decommitment predictions. Models incorporating social media data consistently outperform the baseline model containing only features derived from recruiting and institutional data. In the realm of athletic recruiting, our research can help coaches identify recruits who are more likely to decommit and enable them to proactively adjust recruiting strategies.

论文关键词：NCAA,National Collegiate Athletic Association,NLI,National Letter of Intent,FBS,Football Bowl Subdivision,CART,Classification and Regression Tree,SVM,Support Vector Machine,ANN,Artificial Neural Network,AUC,Area Under Curve,ROC,Receiver Operating Characteristic,Social networks,Twitter,Social media,Predictive analytics,Turnover,College football

论文评审过程：Received 23 February 2018, Revised 11 September 2018, Accepted 8 October 2018, Available online 9 October 2018, Version of Record 17 November 2018.