As a Data scientist for British Airways in a Virtual Experience Program.
Situation: As part of my effort to improve the services of British Airways, I needed to know the insights and gather data on what aspects needed improvement.
Task: To accomplish this, I collected data from the Skytrax reviews website, which had the maximum number of reviews, amounting to around 3100. To make sense of this data, I implemented NLP techniques for better sentiment analysis.
Action: After analyzing the data, I gained a comprehensive understanding of the statistics of the reviews and performed correlation analysis between the different features that affected negative reviews.
Result: My analysis revealed that several aspects of British Airways' services needed improvement. Firstly, the food should be better, particularly for business class. Additionally, the inflight experience needs improvement in terms of comfort, entertainment options, and customer service. Moreover, many customers feel that the services offered by British Airways are not value for money, and the refund process is a hassle, which needs to be addressed. By implementing these changes, British Airways can significantly improve its services, enhancing the customer experience and ensuring customer loyalty.
Situation: As part of the analysis of customer buying behavior, a predictive model was developed using the Xgboost classifier algorithm, and its performance was evaluated.
Task: The goal was to identify the top 5 features that influence customer buying behavior.
Action: The model was trained and evaluated, and the following results were obtained:
Test Accuracy: 85%
AUC score: 0.558
Result: The analysis revealed that the top 5 features that influence customer buying behavior are:
Purchase_led - The duration between date of booking and date of flight departure
Flight Hour
Flight duration
Flight day
Length of stay (for the Round Trip)
The predictive model developed using the Xgboost classifier algorithm demonstrated a good performance with the SMOTE for class minority sampling, achieving an accuracy of 85% on the test dataset. However, the AUC score of 0.558 indicates that the model's ability to distinguish between positive and negative cases is only slightly better than random chance. By considering the top 5 features identified in this analysis, businesses can gain insights into customer buying behavior and tailor their marketing strategies accordingly to improve their sales and customer retention.