St. James Ballroom
Applying Machine Learning and Statistical Methods in the Predicting Client’s Phone Call Activity (303869)
*Xinghe Lu, The Vanguard Group
Chintan Sheth , The Vanguard Group
Keywords: Two-stage model, logistic regression model, random forest, XGBoosting, SVM , cumulative logit model, multinomial logit model
The Vanguard group is an investment advisor based in Malvern, Pennsylvania with over $5.1 trillion in assets under management (AUM). It is one of the largest providers of mutual funds and the second-largest provider of exchange-traded funds (ETFs) in the world.
Vanguard believes in providing superior customer service experience to our clients. Many of our Flagship clients have a dedicated or team of representatives that will assist with their queries and concerns. By anticipating the call activity of our clients, we are able to assign these clients to the right representative. This will ensure that client queries are responded as quickly as possible, thereby decreasing improving customer experience, reducing response delays and assist with representative load sharing.
This poster demonstrates a two-stage modeling approach that was proposed to predict clients’ phone call activity by leveraging both machine learning algorithms and statistical methods, such as Random Forest, XGBoost as well as SVM and Logistic Regression. The team also tried cumulative logit model and multinomial logit models. Python and SAS were mainly used for data preparation, modeling and analysis.