Python libraries: Pandas | Numpy | Matplotlib | Seaborn | Scikit-learn
Machine Learning technique: K Nearest Neighbor(KNN) | Decision Tree | Support Vector Machine (SVM) | Logistic Regression
I used various algorithms to find the best classification method to predict wether a past loan was PAIDOFF or IN COLLECTION. This analysis is the final Capstone Project of the Machine Learning course from my IBM Data Science Certificate. Check out the certificate right here: IBM Machine Learning Certificate
The analysis was performed with the help of:
- Loans dataset provided by IBM.
- Data pre-processing pre-established: dealing with dates data, converting categorical features to numerical values, one hot encoding to set binary variables, normalizing data.
- Scikit-learn metrics available: jaccard score, f1 score, log loss.
- First, I found that KNN was the most accurate method to classify loans of the initial dataset: a K value of 7 provided an accuracy rate of 79% on the test set.
- Then as you can see in the report, SVM appeared to be the most accurate method to classify loans of the new test set (unknown data) with an accuracy rate of 78%%.