Predicting User Purchase Intention in Web Analytics Using Machine Learning with Bootstrapping

The purpose of this research is to investigate the impact of handling class imbalance using Bootstrapping techniques on the performance of various machine learning (ML) algorithms including Random Forest (RF), Decision Tree (DT), and Logistic Regression (LR). Employing Bootstrapping alongside traditional data splitting methods, the study conducts experiments using the Python programming language to assess improvements in model accuracy. The models were trained and tested on the Online Shopper Intention Dataset from UCI's Machine Learning Library, showcasing significant benefits and contributing valuable knowledge to the field of predictive analytics. The findings indicate substantial enhancements, with RF and DT models achieving up to 93.6 % and 89.6 % respectively compared to 89.5% and 86.3% accuracy respectively, While LR reaching 88.01 % compared to 88.2 without bootstrapping. This research not only demonstrates the value of Bootstrapping but also highlights its potential in web analytics. By adopting the recommendations of these experiments, website owners can gain deeper insights into visitor behaviors, offering more targeted services and improving overall website performance.