To develop a model for predicting fraudulent transactions for a financial company and use insights from the model to develop an actionable plan.
Conclusion
We have seen that Accuracy of both Random Forest and Decision Tree is equal, although teh precision of Random Forest is more. In a fraud detection model, Precision is highly important because rather than predicting normal transactions correctly we want Fraud transactions to be predicted correctly and Legit to be left off.If either of the 2 reasons are not fulfiiled we may catch the innocent and leave the culprit. This is also one of the reason why Random Forest and Decision Tree are used unstead of other algorithms.
Also the reason I have chosen this model is because of highly unbalanced dataset (Legit: Fraud :: 99.87:0.13). Random forest makes multiple decision trees which makes it easier (although time taking) for model to understand the data in a simpler way since Decision Tree makes decisions in a boolean way.
Models like XGBoost, Bagging, ANN, and Logistic Regression may give good accuracy but they won't give good precision and recall values