Purchase prediction has an important role for decision-makers in e-commerce to improve consumer experience, provide personalised recommendations and increase revenue. Many works investigated purchase prediction for session logs by analysing users’ behaviour to predict purchase intention after a session has ended. In most cases, e-shoppers prefer to be anonymous while browsing the websites and after a session has ended, identifying users and offering discounts can be challenging. Therefore, after a session ends, predicting purchase intention may not be useful for the e-commerce strategists. In this work, we propose and develop an early purchase prediction framework using advanced machine learning models to investigate how early purchase intention in an ongoing session can be predicted. Since users could be anonymous, this could help to give real-time offers and discounts before the session ends. We use dynamically created session features after each interaction in a session, and propose a utility scoring method to evaluate how early machine learning models can predict the probability of purchase intention. The proposed framework is validated with a real-world dataset. Computational experiments show machine learning models can identify purchase intention early with good performance in terms of Area Under Curve (AUC) score which shows success rate of machine learning models on early purchase prediction.

Purchase prediction can help e-commerce planners plan their stock and personalised offers. Word2Vec is a well-known method to explore word relations in sentences for sentiment analysing by creating vector representation of words. Word2Vec models are used in many works for product recommendations. In this paper, we analyse the effect of item similarities in the sessions in purchase prediction performance. We choose the items from different position of the session, and we derive recommendations from selected items using Word2Vec model. We assess the similarities between items by analysing the number of common recommendations of selected items. We train classification algorithms after we include similarity calculations of the selected items as session features. Computational experiments show that using similarity values of the interacted items in the session improves the performance of purchase prediction in terms of F1 score.

Recommender systems help users to discover and filter new and interesting products based on their preferences. Session-Based Recommender systems are powerful tools for anonymous e-commerce visitors to understand their behaviours and recommend useful products. Diversity in the recommendations is an important parameter due to increasing the opportunity of recommending new and less similar items that users interacted. Effect of diversity has been investigated in many works for the collaborative filtering-based Recommender systems. However, for session-based Recommender systems, exploring the effect of diversity is still an open area. In this paper, we propose an approach to calculate the diversity level of the items in the session logs and analyse the effect of diversity level on the session-based recommendation. In order to test the impact of diversity awareness, we propose a sequential Item-KNN recommendation model. The final recommendation list is created as a contribution of the interacted items in the session that depends on the diversity level between last interacted item of the session. We conduct several experiments to validate our diversity aware model on a real-world dataset. The results show that diversity awareness in the sessions helps to improve the performance of Recommender system in terms of recall and precision evaluation metrics. Also, the proposed method can be applied to other sequential Recommender system methods, including deep-learning based Recommender systems.

Information overloading in e-commerce hinders the consumers' ability to make the right decisions. Customers visiting e-commerce websites can have specific goals in an individual session. However, using sessions that are based on the last item viewed or purchased is not enough to exploit the sessions specific intention or predict users' next actions in the sessions. In this paper, we proposed context and short term user intention aware (CSUI) framework which is based on item similarity collaborative filtering and Association Rule Session-based recommendation systems, the proposed model combines context factor of users' session and users' short term intentions. The developed model has been evaluated on two real-world datasets. Experimental results showed that using session context and users' short term intentions during the recommendation process could help in improving the accuracy of the next item prediction.

No abstract