Some reflections / to do on Final version

[Marijkevandesteene]

For Final version (commit: https://github.com/Marijkevandesteene/MachineLearning/commit/f2b9feba83d2c51d6092343e4c30191b413604a5) some final thoughts / to do:

• [ ] generate html

• [ ] add csv with final selection to github / mail to Bart

• [x] executive summary: start with figures of gain? add figures of actual gain on test set?

• [x] I believe there is still a windows path used in the notebook

-- Needed for explainability: Mind to use the correct target for the corresponding model

train_V2 = pd.read_csv('.\output\train_v2_prep_without_outliers.csv') score = pd.read_csv('.\output\score_prep_without_outliers.csv')

• [x] scaling needed for linear surrogate explainability (SHAP) - see conclusion?

• [x] discussion on scoring model classifying what clients cause damage

• [x] conclusions model selection? to mention?

1. [x] performance models
2. [x] data preparation?
3. [ ]

[dluts]

Data prep and performance of the models is done.

[dluts]

exec summary and model selection is done

[dluts]

I thin we will leave the scaling be for shap?