Gemini

In an era characterized by rapid technological advancements and increasingly complex socio-technical systems, especially regarding Artificial Intelligence (AI) based E-Commerce Optimization, the demand for a comprehensive method to resolve problems, which could result from the lack of trust in AI-Software-Systems, has never been more significant. Systems thinking offers a robust framework for understanding the intricate relationships between various components of a system, making it particularly relevant in today’s digital and AI-based software landscape. Concurrently, the E-Commerce sector is undergoing a transformation, driven by advancements in AI and data analytics grounded in legal changes regarding the General Data Protection Regulation (GDPR) and the upcoming EU AI Act. We explore the synergies between AI-based E-Commerce and Systems Thinking by introducing the AI-DATA model, a human-centered Customer Journey Optimization approach in the context of E-Commerce marketing strategy based on Trustworthy AI. AI-DATA is a phase model consisting of Awareness, Interest, Desire, Action, Trust and Again, and is an extension of the AIDA model, which also considers the customer path to contain activities after a transaction has finished, and a cycle of repeat customers.

In the quest for improving systems, understanding user behaviour is a crucial step towards achieving optimal results. One technique that aids in comprehending user behaviour is process mining, which involves analysing event logs of a system. However, the accuracy of the process mining results can be impeded by large amounts of data or very unstructured processes, leading to confusion and difficulties in interpretation. To overcome this challenge, preprocessing techniques exist to enhance the resulting process maps. This work proposes a preprocessing method that utilizes an event hierarchy to produce a sanitized process map, that is easier to interpret. The proposed method involves analysing the event logs and creating a hierarchical structure of the events. This hierarchy allows the identification of relevant events and their corresponding dependencies, which in turn enables the extraction of useful process information. This preprocessing method has been tested on event data from an online newspaper, and it has shown promising results in improving readability of the resulting process maps. By utilizing an event hierarchy, relevant information can be extracted from large amounts of data and unstructured processes, leading to cleaner process maps that aid in understanding user behaviour.

This paper presents an ontology and process mining-based system that employs semantic technologies for targeted e-commerce marketing strategy suggestions. The proposed approach utilizes ontologies as a knowledge representation formalism to capture the domain knowledge, product descriptions, and website texts in the e-commerce domain. The ontology is populated with relevant attributes and corresponding values, allowing for a structured and standardized representation of users and content. To suggest targeted marketing strategies, the system employs interest and intensity calculation techniques that utilize the concept of scroll depth and others to measure the interest in different contents and customer preferences. Process mining techniques identify moments of truth along the customer journey and happy paths a customer usually goes through. Once the ontology is combined with the process mining results, it can be identified in which phase of the customer journey a user is, and it can be determined what marketing actions to employ to get them to enter the next phase of a marketing model. Overall, the proposed approach is a promising approach to address the challenges of personalized e-commerce marketing strategy suggestions by utilizing ontology-based knowledge representation and intensity calculation techniques. The system has the potential to provide personalized and targeted marketing strategies for e-commerce businesses, leading to increased customer engagement, loyalty, and sales.