Fraud Detection

Goal We aim to explore the options when we have highly skewed dataset to prepare the model and best performance meteric to be used for evaluation and prepare a model comparison report. Work Plan: What performance metric we should choose and Why? How to deal with highly skewed dataset - Undersampling , Oversampling , SMOTE Perform Precision-Recall trade off Go to Github source code

March 9, 2020

Customer Purchase Behaviour Study

Overview We need to understand our customer’s purchase behaviour and rank them on the basis of how much they spend, how frequent they visit and how recent they visited. Research shown: Cohort Analysis Customer Retention Recency Frequency Monetary Model (RFM) EDA Elbow method for Optimum clusters Go to Github source code