Newsletter Subscription Predictor

Project Overview

This project investigates player engagement within a Minecraft environment to predict whether a user will subscribe to a game-related newsletter. By analyzing Age and Played Hours, we built a classification model to help marketing teams target potential subscribers more effectively.

Methodology

Using the tidymodels framework in R, the analysis followed a rigorous data science workflow:

• Data Wrangling: Filtered quantitative predictors and converted subscription status into factor variables.
• Pre-processing: Centered and scaled variables to ensure distance-based calculations were not biased by unit differences.
• Model Tuning: Performed 5-fold cross-validation testing K-values from 1 to 100 to identify the peak accuracy threshold.
• Evaluation: Utilized a confusion matrix to calculate precision and recall, identifying a high propensity for capturing true positives.

Key Insights

The model demonstrated that played hours and age are adequate exploratory variables for predicting player behavior. A significant finding was the exceptionally high recall (97.2%), suggesting the model is highly effective at identifying nearly all potential subscribers, making it a valuable tool for engagement-driven marketing strategies.