Predicting Credit Risk with Supervised Machine Learning

             _._._                       _._._
            _|   |_                     _|   |_
            | ... |_._._._._._._._._._._| ... |
            | ||| |  o NATIONAL BANK o  | ||| |
            | """ |  """    """    """  | """ |
       ())  |[-|-]| [-|-]  [-|-]  [-|-] |[-|-]|  ())
      (())) |     |---------------------|     | (()))
     (())())| """ |  """    """    """  | """ |(())())
     (()))()|[-|-]|  :::   .-"-.   :::  |[-|-]|(()))()
     ()))(()|     | |~|~|  |_|_|  |~|~| |     |()))(()
        ||  |_____|_|_|_|__|_|_|__|_|_|_|_____|  ||
    ~ ~^^ @@@@@@@@@@@@@@/=======\@@@@@@@@@@@@@@ ^^~ ~
        ^~^~                                ~^~^

---

Tools, Languages, & Libraries Utilized

Python

Pandas

Pathlib

Sklearn - LogisticRegression, RandomForestClassifier

VS Code

Jupyter Notebook

---

A machine learning model that attempts to predict whether a loan from LendingClub will become high risk or not.

Credit Risk Evaluator

Which model will be better? LogisticRegression or RandomForestClassifier?

Predictions:

I predict that the RandomForestClassifier will be a better model due to the data set containing several columns of categorical data.

Thoughts before scaling data

Comparisons from unscaled data:

The logistic regression model seems to be performing better on the unscaled data. The random forest classifier is performing poorly, possibly due to overfitting.

Predictions for scaled data:

I think the performance of the unscaled data is too poor in the random forest classifier to overcome the deficit of the logistical regression model. I predict both models will perform better after scaling.

Results

Comparisons from scaled data:

It seems the prediction was incorrect due to random forest classifier actually performing worse after scaling. The logistic regression model performed significantly better after scaling.

Results for scaled data:

The logistic regression model seems to be the way to go meaning my prediction was incorrect. The data set could possibly use some more pre-processing to achieve better results.