Credit default payments
Budget: $30 – $250 USD
budget of 150$
On the dataset:
Can we predict if a customer will default on payments or not?
Using range of models including:
- Fit a logisitic regression model to this data see which solver is best– need to transform to log some of the columns for better results. What is the accuracy score?
o Reduce the dimensionality of the dataset using PCA, any better results?
- Run KNN model
o Same - Reduce the dimensionality of the dataset using PCA, any better results?
- SVM
- XGboost
- And ultimately neural networks – this should be the best
We want highest possible recall score, from classification report
Plot train vs test data accuracy for each model, and ROC curve to see
Optimize hyperparameters to get the best results - Use gridsearch , cross validation
Redo models
Observations found i.e., which demographic most likely to default
On the dataset:
Can we predict if a customer will default on payments or not?
Using range of models including:
- Fit a logisitic regression model to this data see which solver is best– need to transform to log some of the columns for better results. What is the accuracy score?
o Reduce the dimensionality of the dataset using PCA, any better results?
- Run KNN model
o Same - Reduce the dimensionality of the dataset using PCA, any better results?
- SVM
- XGboost
- And ultimately neural networks – this should be the best
We want highest possible recall score, from classification report
Plot train vs test data accuracy for each model, and ROC curve to see
Optimize hyperparameters to get the best results - Use gridsearch , cross validation
Redo models
Observations found i.e., which demographic most likely to default