Machine Learning for Feature Importance: Random Forest Classification on Service Satisfaction

Project Name

ForestFlight

Problem

Strict budget limits and "one-size-fits-all" strategic marketing prevent airlines from maximizing passenger satisfaction. Without segment-specific data, it is impossible to prioritize the strategic investments that best address the unique needs of different generations and flight classes.

Objective

Identifying the top service drivers by classifying satisfaction across flight classes and generational segments to guide strategic investment and marketing activities.

Data

Binary classification of Overall Satisfaction (Satisfied vs. Not Satisfied).

Customer ratings for on-board services (Seat Comfort, Legroom, Wi-Fi) and ground services (Booking, Baggage, Gate).

Categorical data for Flight Class (Economy/Business) and Demographics (Gen Z, Millennial, Gen X, Baby Boomer).

Method used

Random forest classification Model

Results

Mapped the top 6 service priorities for Economy and Business class, revealing that service importance shifts significantly based on the service tier.

Uncovered a generational divide in Business Class, where Gen X passengers prioritize physical comfort (Legroom and Seat Comfort), while Millennials and Gen Z prioritize digital connectivity (In-flight Wi-Fi and Entertainment).

Confirmed uniform preferences in Economy Class, finding no significant variance between generations; this allows for a streamlined, universal strategic investment approach for the Economy cabin.

Tools

Python - Tableau

Project Slides

Download

Python data modeling: Preprocessing airline passenger demographics and feature engineering using Pandas.

Random Forest classification in Python: Machine learning model training and feature importance extraction.

Random Forest analysis: Identifying top satisfaction drivers for Economy Class using machine learning.

Segmented data modeling: Random Forest feature importance for Gen X Business Class satisfaction.

Machine learning visualization: Donut chart showing significant feature distribution in satisfaction models.

Deep Into Project

The ForestFlight project uses a "Wisdom of the Crowd" approach to analyze customer feedback. Instead of relying on a single calculation, the system builds hundreds of individual "decision models" that work together to predict whether a passenger will leave satisfied. By looking at how these models make their choices, we can measure the "Feature Importance"—essentially a ranking that tells us which specific services (like Wi-Fi or Seat Comfort) are the most powerful triggers for customer loyalty.

To make these insights actionable, I divided the data into specific groups based on age and travel class. This allowed the system to see patterns that a general overview would miss. For example, it revealed that what makes a flight "successful" for a young professional in Business Class is completely different from what a frequent flyer in Economy values. By pinpointing these specific needs, the project provides a clear roadmap for where the airline should spend its budget to get the best possible results from its strategic investments and marketing.