Customer Propensity Model: Marketing Campaign

A marketing campaign being conducted undeniably consumes lots of resources and funds to be carried on. It is vital to ensure that such campaigns being conducted are effective and the resources used are not going in vain. Thus, it is important to recognise potential customers who might respond to the campaign and shrink the range of the target market. This research considers four different models, Logistic Regression (LR), Random Forest (RF), Extreme Gradient Boost (XGBoost) and Extreme Gradient Boost Random Forest (XGBRF) to determine the best model for a propensity model on a marketing campaign dataset. Exploratory Data Analysis (EDA) along with Feature Engineering were conducted to prepare the dataset, understand the as well as to generate new useful features for building the model. The best model was selected by evaluating them on their accuracy, precision, recall and F1-score. Once the metrics are obtained, the evaluation is repeated after tuning the hyperparameters. XGBoost was found to be the best model in building a propensity model where the hyperparameter tuning was less impactful.