This project looks to identify what insights can be drawn from the public's sentiment towards 6 different US airlines. Tweets.zip contains the tweets used in this project.
Each of the techniques have been developed using Google Colaboratory therefore all required dependencies are included in the ipynb files.
Modelling:
- Do you notice anything about the data from exploratory analysis?
- Can you form a ‘baseline to beat’ by training a simple, naive model to begin with?
- How would neural networks perform vs non-neural network models?
- How well can pre-trained models perform on the dataset?
- When selecting a model in a business environment, what factors would you consider apart from predictive performance (e.g. accuracy)?
Insights:
- What are the key predictors of your model that are most important in the model’s classification?
- What were the challenges? Are there any caveats or risks with interpreting outputs?
- What can we understand about sentiment towards airlines?
- What technical next steps would you suggest if this was to be continued?
- What business next steps would you suggest based on current findings?
Authors: Victoria Porter, Will Holbrook, Laura Cope, Lauren Cooper and Joel Wolinsky