Creative Commons

E-commerce has gained significant popularity and this trend is expected to grow even more. The users do not always visit an e-commerce with purchase intention but with browsing one meaning that they interact with the products but ultimately abandon the website without offer revenue to the business. This is a challenge that any e-commerce faces. Users interacting with it generate a huge volume of data that business can benefit from to convert browsers into buyers and simultaneously to adjust the marketing strategies accordingly with aim to improve customer experience. Understanding customer intent in real-time is considered vital to improve data-driven decisions. Machine learning offers the framework to build models that can identify the objective of a user. Many studies have utilized it and managed to predict purchase intention with high accuracy. None of them took the advantage of online machine learning that allows the models to be updated continuously without retraining needed. This study utilized probabilistic, linear and tree-based online machine learning methods with goal to achieve this, using a well-known experimental dataset from UCI ML repository. In addition, the impact of features dimensionality on the models’ performance examined. The models evaluated in terms of AUC, sensitivity and specificity and the results suggested that online classifiers are considered promising to achieve this task, with tree-based model pointing out an overall better performance.