
Table of Content
Problem Statement
Our airline client was struggling to increase their customer base and wanted to find a way to incentivize current customers to refer new passengers. They had tried various referral programs in the past, but none of them had been successful in increasing their customer base.
After conducting research, we found that the main issue with their previous referral programs was that they were not targeted towards the right customers. The airline needed a way to identify which customers were most likely to refer new passengers and target them with personalized referral incentives.
Objective
We developed an AI-powered predictive model that analyzed customer data such as travel history, booking frequency, and loyalty program participation to identify which customers were most likely to refer new passengers. We then targeted these high-value customers with personalized referral incentives such as discounted flights or bonus loyalty points for each new passenger they referred.
Dataset
Data Pipeline
Project Workflow

Code (github)
Data Visualization
Conclusion
The Models used for this Classsification problem are
Logistic Regression Model
Decision Tree Model
Random Forest Model
K-Nearest Neighbor Model
Support Vector Machine Model
Naive Bayes
We performed Hyperparameter tuning using Gridsearch CV method for Decision Tree Model, Random Forest Model , K-Nearest Neighbor ,Support Vector Machine and Naive Bayes. To increase accuracy and avoid Overfitting Criteria, this is done. After that, we finalized the Gradient Boosting model by fine-tuning the hyperparameters.
Based on the knowledge of the business and the problem usecase. The Classification metrics of Recall is given first priority , Accuray is given second priority , and ROC AUC is given third priority.
We have built classifier models using 6 different types of classifiers and all these are able to give accuracy of more than 90%.* We can conclude that LogisticRegression gives the best model.
model evaluation metrics comparison, we can see that Support Vector Machine being the model with highest accuracy rate by a very small margin, works best among the experimented models for the given dataset.
The most important feature are overall rating and Value for money that contribute to a model's prediction whether a passenger will recommened a particular airline to his/her friends.
The classifier models developed can be used to predict passenger referral as it will give airlines ability to identify impactful passengers who can help in bringing more revenues.
As a result, in order to increase their business or grow, our client must provide excellent cabin service, ground service, food beverage entertainment, and seat comfort.
Skills
Data AnalysisData VisualizationMachine LearningPython