Predictive Modeling for Marketing Campaign Efficiency

This project aimed to predict customer responses to marketing offers, enhancing the efficiency of future campaigns. Through data analysis and predictive modeling, key variables influencing customer response were identified, providing valuable insights for targeted marketing strategies.

See the code

Step-by-Step Process

Understanding the Dataset

2240 rows and 29 columns.
The dataset consisted of various variables such as campaign acceptance history, customer demographics, purchasing behavior, and complaints.
The target variable was "Response," indicating whether the customer accepted the offer in the last campaign.
Found mean, standard deviation and other statistical information of the variables in the dataset.

Data Cleaning and Preparation

Removed missing values.
Removed nonsensical and unnecessary data.
Created additional variables to better represent the data (Total_Purchases, Total_Spent, Age_Range, etc.)

Exploratory Data Analysis

Uncovered patterns, relationships, and gained insights that will guide the subsequent modeling process and final recommendations.
Created graphs and visualizations of distributions such as age, income, education, marital status, total spent, total number of purchases, etc.
Built a correlation matrix and discovered the relationships between the variables.

Model Selection, Training, and Evaluation

Selected appropriate machine learning models for binary classification, such as logistic regression, decision tree, and xgboost.
Split the dataset into training and testing sets for model evaluation.
Trained the models and observed the results with the accompanying confusion matrices.
Found that xgboost performed the best with accuracy: 0.88, precision: 0.67, recall: 0.42, and f1 score: 0.52.

Recommendation

According to the xgboost, the most significant variables to consider are Total_Offers, AcceptedCmp3, In Relationship, Seniority, NumWebVisitsMonth, and Recency.
After thorough analysis and testing, my recommendation for the next marketing campaign was to focus on individuals who have previously accepted offers, have longer enrollment tenure with the company, recently made purchases, are single, and have demonstrated a high number of visits to the company website in the past month.