The report was very fluid to read, and was complemented with lots of helpful visuals. The problem is well explained and the data that you used was rigorously cleaned, with good amounts of attention to potentially fraudulent sales. You were very persistent on tackling the messy dataset from many angles, and I think that was well reflected within the report.
I would have liked to see some more deliberation behind the selection of features. Maybe mention why you chose to filter out the ones that did not make it into the modeling process? Also, what were the reasons behind choosing the models as such, and why you think l1 regularization worked the best? It seemed a little arbitrary here in the report.
Overall, great project and I hope it was a good experience!
The report was very fluid to read, and was complemented with lots of helpful visuals. The problem is well explained and the data that you used was rigorously cleaned, with good amounts of attention to potentially fraudulent sales. You were very persistent on tackling the messy dataset from many angles, and I think that was well reflected within the report.
I would have liked to see some more deliberation behind the selection of features. Maybe mention why you chose to filter out the ones that did not make it into the modeling process? Also, what were the reasons behind choosing the models as such, and why you think l1 regularization worked the best? It seemed a little arbitrary here in the report.
Overall, great project and I hope it was a good experience!