British Airways Good or Bad?

This project is a part of the Data Science virtual internship program offered by Forage with British Airways.

The virtual Internship is divided into two main tasks

1. Web scraping to gain company insights
2. Predicting customer buying behaviour

Task 1 - Web scraping to gain company insights

Customers who book a flight with BA will experience many interaction points with the BA brand. Understanding a customer's feelings, needs, and feedback is crucial for any business, including BA.

This first task is focused on scraping and collecting customer feedback and reviewing data from a third-party source and analysing this data to present any insights you may uncover.

Customer review data for Britis Airways was collected from Skytrax.

Data Cleaning & Exploration:

Reviews were cleaned for punctuation, spelling, and special characters.

Exploratory Data Analysis (EDA) revealed the following:

• Average Overall Rating: The average rating for the reviews is 4.58, indicating a generally positive overall customer sentiment.
• Top 5 Reviewing Countries: UK,USA,Australia,Canada,Germany
• Highest Rating Country: Dominician Republic,Ecuador,Costa Rica,Japan.
• Periods of Decreased Ratings:From March 2020 to October 2021, there was a decrease in reviews due to travel restrictions caused by the Covid pandemic.

Text Analysis:

• Wordcloud visualization identified the most frequent keywords: flight, seat, service, cabin crew, and good experiences. However, negative sentiments related to delays, problems, and bad experiences are also evident.
• N-gram analysis revealed positive sentiment towards cabin crew.
• Vader sentiment analysis indicated 55.5% positive reviews.

Following insights were uncovered as they are summed up in the one slide presentation.

Task 2 - Predicting customer buying behaviour

Objective:

Predict customer buying behavior for flight bookings.

Methodology:

• Model Training: Two models were trained: Random Forest Classifier and XGBoost Classifier.
• Model Evaluation: F1-score was used as the primary evaluation metric.
• Class Imbalance Handling: Oversampling and undersampling techniques (SMOTE-RUS) were employed to address the class imbalance in the dataset.
• Feature Importance: XGBoost feature importance was analyzed to identify key factors influencing bookings.

This task involves building a high quality predictive to predict the successful bookings using customer bookings data.

Conclusion:

The predictive model successfully identified key factors influencing customer flight bookings, such as sales channel, trip type, length of stay, and passenger count. By leveraging these insights, British Airways can:

• Target marketing efforts more effectively.
• Personalize recommendations for each customer.
• Optimize flight pricing dynamically.
• Proactively engage potential customers.

These actions can lead to increased flight bookings and improved customer satisfaction.