PiotMik
commented
2 years ago Some ideas as a reference point: https://rpubs.com/swapnil_s_patil/binaryClassification
Most popular in practice are the ones covered on the lectures:
Additionally from the ones not covered some popular choices are:
Read about binary classification models and prepare a choice of 2-4 interesting ones to try on the project dataset. Keep in mind the following guidelines:
There should be at least one "baseline" model, which is super simple and is used mainly to assess how well can a naive model perform to understand if the more complicated ones actually are doing a relatively good job.
Then you can look up 1/2 models which are more interesting. Here, because of the topic chosen (not only predict Satisfied/Not Satisfied, but also help the Airlines improve the most relevant services) we should aim for models which are explainable, that is you can actually say how the model goes from inputs to prediction. E.g. Airlines should improve the quality of seats, because as per logistic regression this variable has the biggest negative contribution to the overall satisfaction.
Optional: Take one state-of-art model and see what is actually a max achievable accuracy for this dataset
[x] Propose models
[x] Select models
[x] Create& assign further tasks accordingly