A Data-Driven Structure for Targeted Advertising and Digital Marketing Analytics: Intelligent Customer Segmentation

Abstract

 In order to encourage the use of targeted advertising and digital marketing in the existing 
online marketplaces, this study proposes a clever, data-driven client segmentation strategy. In order 
to identify significant consumer groups using behavioural, demographic, and psychographic 
variables, the framework combines machine learning algorithms, such as Principal Component 
Analysis (PCA) to reduce dimensionalities, K-Means and DBSCAN to perform unsupervised 
clustering, and Gradient Boosting Machines (GBM) to predict segment-level responses. A dataset of 
50,000 customer records from a digital retail environment was used to evaluate the model.  
Data preparation, feature engineering, and multicollinearity reduction using PCA—which has 
preserved 92% of the variance in 15 transformed components—are the first steps in the technique. 
When K-Means and DBSCAN were tested for segmentation quality, the former (k=5) performed 
better according to the Davies-Bouldin Index (0.42) and Silhouette Score (0.71). There were some 
discernible behavioural variations in the frequency of purchases, browsing habits, campaign 
responses, and spending across the five consumer groups that were subsequently created. With an 
F1-score of 0.87 and an AUC of 0.93, a GBM classifier to evaluate the impact of marketing was also 
trained to predict the probability of ad-clicks in clusters.