Lending Club is a peer to peer lending company based in the United States, in which investors provide funds for potential borrowers and investors earn a profit depending on the risk they take (the borrowers credit score).
Build a pipeline to process the data and predict if a new loan or increase should be approved for those customers who are already on profile
Data has 145 features which contains 109 numeric columns and 36 categorical columns. Manually looked through the data and dropped some features which obviously have no relationship with the target variable based on our research. Some features are relating to the target variable but they have too many NAs so we assigned ‘None’ to categorical variables and 0 or -1 to some numerics features and other numeric features were assigned to mode value.
We built three models: logistic Regression, Random Forest Regression, and Neural Network. Details about metrics to measure the performance are in the PPT file and code file. Logistic regression has the best performance with accuracy 95%, precision 91% and recall 99%. High recall means positive instances are correctly detected by the calssifier that would be great because we don't want to miss any qualified client.
https://drive.google.com/open?id=1hx6mGt1m_Gf2XQyVDqAFIkYU05gT0MXT
saved models were too big to upload as well, please ignore code that loads saved models and retrain.
Best Model - LOGISTIC REGRESSION Neural Network doesn’t seem to generalize well to test data but there’s a lot of room for improvement/tweaks to get there. Based on the metrics, We can conclude that loans predicted by the random forest as bad loans are either bad or have a high possibility of becoming so. For more details, please look through the power point file, consulting proposal.