Task 1:
Scrape and analyse customer review data to uncover findings for British Airways.
Task 2:
Build a predictive model to understand factors that influence buying behaviour.
First data was scrapped from Skytrax website.
The next stages involved cleaning the data and checking for missing values. Further details on this can be seen in the notebook.
How have overall customer ratings for British Airways evolved over time?
Observation
- After decomposing the time series additively, we see a clear and consistent decreasing trend in the overall rating from customers over the period of 2015 to 2023.
- The residuals are centred around 0, which supports the use of an additive model.
- The noise observed from mid-2020 to mid-2021 in the residuals may be attributed to the linear interpolation used to fill in the missing values during that period.
Distribution of review ratings for British Airways
Observation
- Looking at the distribution for overall rating, we see that most people give a rating of 1. Other than that, there does not seem to be any difference between the other ratings.
- We see that most guests are not satisfied with the WiFi service.
- The staff service receives high ratings, with most people giving 4 and 5 stars.
- There seems to be a uniform distribution of seating comfort.
- Most people write between 500 and 1000 words per review.
Length of reviews per travel class
Observation
- We see, there does seem to be an unbalanced dataset. A lot more people have travelled to the economy and business section of this dataset. The distributions follow the same shape between different travel classes.
Overall rating is given per travel class
Average overall rating per travel class
How many customers would recommend British Airways?
Word Cloud. What are the most frequent words?
Performance Metrics
| Metric | Value |
|---|---|
| Accuracy | 0.846 |
| Precision | 0.483 |
| Recall | 0.123 |
| F1-score | 0.197 |
The model has a high accuracy but a low precision and recall. So more work can be done to improve it.