This repository is an attempted solution to the dasa case study for data scientist.
It does not follow the guidelines 100%. (I apologize for that)
It has two main directories; data and notebooks.
Data contains the original datasets as well as a manufactured data set.
Notebooks contain four different ipynb files. (standard libraries from anaconda python 3.6 as well as lightgbm)
(didnt create a requirements file as well - sorry)
The notebooks directory contains these files:
- EDA
- FE
- Model Evaluation
- Eval Results
They should be seem in that order. Eval results can be run (given a path is provided) to evaluate the model on a test set.
Other notebooks are well commented and should be self explanatory.
Further improvement opportunities would be:
- Feature Selection
- Outlier removal for trainning
- Further feature engineering
- Gather more data about events (ideally patient & doctor meta information)
Any doubts or suggestions
ps: its important to notice (as noted in the FE.ipynb file) that the target variable is not the given "duration" but rather a value that has to be calculated.