From Clicks to Conversions: Predicting Online Purchase Intent Using Supervised Machine Learning

Abstract

The rise of the E-Commerce sector has transformed the brick-and-mortar businesses to a new paradigm by enabling businesses to gain access to a wider local and international market, reach out to diversified customer groups and allow effective sales and distribution channels for their products and services. Despite the benefits e-commerce brought, many businesses struggle to extract meaningful insights from users data to improve business strategies and increase customers' conversion rates. Hence, this study aims to analyze customers browsing behaviours to predict whether a user session could potentially lead to a successful purchase. A comparative analysis will be performed using the “Online Shoppers Purchasing Intention” dataset on Decision Tree, SVC, Random Forest and XGBoost to identify the best classifier in making predictions. The findings indicated that Random Forest and XGBoost had achieved F1-score of 0.66 and AUPR of 0.71 and 0.7, indicating a strong predictive balance and ability in identifying likely purchasers. PageValues, BounceRates, ProductRelatedDuration and ExitRates were found to be strong predictor variables. The findings are vital in helping businesses to design better business strategies that could offer promotions to the target group that has high probability in purchasing and hence increase conversion rates and business revenue.