Student researcher uses SAS to predict user ratings and recommendations of airlines

1

The holidays are always busy when it comes to travel. Finding a place to park and long security lines at the airport are all expected when you’re flying around the holidays. This week we hear from a student who conducted research which looked at customer satisfaction for popular airlines and how that relates to customer retention.

Parijat Ghosh

Student researcher uses SAS to analyze airline data

I am Parijat Ghosh and I am from India. I am a graduate student at Oklahoma State University where I’m pursuing my masters in Agriculture with a minor in Information Systems. I have used SAS for two years and have attended several SAS conferences. As a student researcher, I enjoy contributing to the SAS community by presenting my research. Most recently I presented a poster at Analytics 2012. My current work deals with Predicting User Ratings and Recommendations for Airlines.

Customer attrition is a serious concern in the world of business. If airlines do not pay attention to the promises they have made to their customers, this can result in customer churn. Customer churn is a common concern in several industries. Much time is spent on trying to analyze the factors that cause churn. Rather than focus on churn and factors that are involved, my approach for this project was from the flip side. I chose to focus on customer retention. My belief is that if airlines meet the expectations they have established with their customers, then customer retention would be an inevitable thing.

You may ask why I focused on customer retention. It can be more costly to gain the trust of a new customer than it is to maintain a current customer. This is why it is important for companies to focus on their current customers and not just put their efforts into gaining new customers. This is what made me interested and led to my research.

The objective of the research study was to analyze user ratings of different airlines and build a predictive model to predict these types of ratings. This research study also examined the driving forces that compel customers to be loyal. SAS Enterprise Miner was used for data exploration and for building the predictive model. Neural Network (RBF and Multilayer Perceptron), Decision Trees (Gini, autonomous and entropy) and Regression (polynomial) followed by a model comparison were used. The two target variables were overall user rating (10-point scale) and recommendation (binary variable – yes/no). User ratings were for specific aspects of the flight such as value for money, seat comfort, staff service, catering and entertainment. There is also a nominal variable for the class of service flown (economy, business, etc.) The model comparison node selected MLP Neural Network as the best model for both the target variables. Because Neural Network is opaque with respect to importance of input variables, a surrogate decision tree was used to analyze the predictions from the Neural Network. The common theme across the two surrogate decision trees was seat comfort, staff service, entertainment and value for money. The surrogate decision tree results provided insight into which variables were driving the target. For more information on this research, you can find additional details in the paper itself .

Thanks for sharing your research Parijat! If you are a student interested in learning more about SAS, checkout the SAS Student Program website.

Share

About Author

Julie Petlick

SAS Student Programs Manager

Julie Petlick works in the SAS Education Division as part of the Global Academic Program team. Julie is responsible for the SAS Student Program and is dedicated to supporting teaching and learning.

Related Posts

Leave A Reply

Back to Top