To provide a bundle of high-value services and more specifically, smart personalized and quick retail recommendations at the edge for a large number of users using historical data and machine learning techniques, while combining the functionalities of two systems that comprise 5GPACE (i.e. the Immersive Media Services provided by our partner ITALTEL with Machine Learning-based personal retail recommendations offered by Incelligent).
Existing machine learning schemes were designed and adjusted to deliver recommendations at the edge and to be regularly updated centrally. Through the microservice (“Service Mesh”) paradigm Incelligent and ITALTEL were able to seamlessly integrate their functionalities into a unique framework. This way, interworking processes between systems and applications are greatly simplified and more easily deployed by avoiding the development of proprietary interaction mechanisms.
During the course of the MATILDA (http://www.matilda-5g.eu/) phase 2 5G-PPP project, partners have designed and developed a holistic MATILDA framework that supports the development of 5G-ready applications, the creation of the on-demand required networking and computational infrastructure and the activation of the appropriate networking mechanisms for the support of the industry vertical applications. Our solution was demonstrated on top of the MATILDA platform but can be deployed easily elsewhere due to its microservice design approach. Furthermore, in-house/ proprietary ML recommendation schemes were employed.
Delivering Personalized Retail Recommendations to Users When in a Mall, a Touristic Location and/or Market Area
DESCRIPTION WHAT WE HAVE DONE/ACHIEVED
5GPACE has been deployed in a real environment (CNIT premises in Genoa), although in smaller scale, reaching a TRL level 5/6, in the context of the MATILDA EU project. During the final project review, ITALTEL and Incelligent were able to demonstrate the 5GPACE functionalities, i.e. the high resolution video sharing capabilities and the provision of Machine Learning-based recommendations, on top of the MATILDA platform and therefore proving the applicability of the use case. The scenario involved the following: All target locations have working Bluetooth beacons. User A goes to the designated Location 1 where he/she gets a recommendation for the next shop. 5GPACE recommends him to visit in the form of a video which is streamed to his device. Location 1 has two possible videos, e.g. a clothing store and a shoes store video. User A then visits Location 2 and gets a new recommendation for e.g. a Grocery store or a Coffee place.