To optimize personalized offers and reduce cart abandonment, it is essential to understand customer behavior in e-commerce after products are added to the cart. Although purchase prediction models are well researched, session-level changes, including price variations, product category shifts, and geographical context, are less examined concerning their impact on machine learning models for predicting purchase behavior after cart additions. This study incorporates these factors into machine learning models to examine their impacts on predictions using explainable AI techniques. The comprehensive experimental results obtained from two datasets and eight models demonstrate that machine learning algorithms can achieve an F1 score of 89% in predicting purchase behavior following cart additions. This study highlights the significant impact of session-specific factors, like price fluctuations, category transitions, and geographical context, coupled with consumers’ previous browsing patterns, on model predictions.

Purpose: Customers' participation in the online health community to create value with the brand is growing research interest. In addition, customers are using social media platforms to create value in the food sector. This rises points to the need to study consumers' interactions with online communities and the role of social media content and customer satisfaction in such an environment.Design/methodology/approach: This research collects data using a survey approach. The data were analysed using a partial least square-structural equation modelling (PLS-SEM).Findings: The findings indicated the impact of social media content and satisfaction on value co-creation in healthy food online communities. The study’s results provide significant new insights into the food sector during the pandemic.Originality/value: This research enhances the knowledge of satisfaction and value co-creation in the social media context. The findings build on the previous literature on value co-creation, add to the food sector and explain the mediating role of satisfaction between social media content and value co-creation.

The reputation of a business is significantly influenced by online reviews, with negative feedback having the potential to harm a brand's image and dissuade potential customers. To safeguard their image and convert dissatisfied users into loyal ones, businesses must formulate effective strategies for managing negative reviews. This study investigates response strategies aimed at enhancing the relationship between people and organizations among dissatisfied users upon their return. Using AI as a methodology by leveraging machine learning in our research, we managed to achieve remarkable accuracy using only response attributes to predict there is an increase in subsequent ratings of dissatisfied return customers. The study reveals that specific actions taken or planned in response to a user's complaint, a statement accepting responsibility for service failures, and a request for direct contact through phone or email can positively impact user loyalty and elevate subsequent ratings from returning dissatisfied customers. However, there is a noteworthy negative correlation between the length of the response text and the subsequent rating from returning customers. These findings not only provide theoretical insights but also have practical implications, underscoring the value of machine learning and data analytics in effective reputation management.

E-commerce prediction models may lose reliability when data from heterogeneous websites are pooled indiscriminately. We develop LACAT, a leakage-aware, contextadaptive transfer framework for three-class consumer-journey prediction across seven e-commerce websites comprising 5,558,318 raw interactions. LACAT combines website-specific source experts, strictly prior-session behavioural histories, limitedlabel target adaptation, and a reliability gate based on protected target validation. Under matched training budgets, source experts outperformed pooled models, especially for transaction ranking and probability quality. Leakage-safe history improved prediction across all websites, while gated adaptation further improved probability quality. The findings support context-preserving, temporally valid, evidencecontingent cross-website analytics without assuming pooled scale is beneficial.