Constructive recommendation systems for retailers

Retailers love to identify customers' needs so they can respond to them. The newly opened Industrial Doctorate position is about helping in this. The Doctorate aims at developing a new generation of recommendation systems, leveraging on the larger data sets that are becoming available and leveraging on the increased product or service flexibility resulting from the Industry 4.0 revolution.

Applicants should have an interest in the application of machine learning and recommendation system techniques. The Industrial Doctorate is a cooperation between the EIT Digital Doctoral School, The University of Trento and Telecom Italia (TIM).

Challenge

In retail, the capability to better address customer individual needs is getting critical given the offer explosion. Retailers should be able to provide the most appropriate set of pricing and product/service alternatives flexibly intercepting customers' needs and offer a richer collection of services. At the same time, it should be easier for the customers to engage with different providers to meet all their needs.

In retail, the capability to better address customer individual needs is getting critical given the offer explosion. Retailers should be able to provide the most appropriate set of pricing and product/service alternatives flexibly intercepting customers' needs and offer a richer collection of services. At the same time, it should be easier for the customers to engage with different providers to meet all their needs.

Unfortunately, most of the existing recommendation systems focus on some solutions/items that a user may like but fail short in fulfilling the goals of retailers. Traditional recommendation systems are either based on:

1. content-based recommendation: how the features of the candidate items match the user model,
2. collaborative filtering: how similar users liked/chose the same candidate items, or
3. a combination of 1 and 2.

In line with the Industry 4.0 paradigm, the interaction with the customer should be usable for packaging the product or service offering and this is what future recommendation systems should provide. Increased flexibility in the production and delivery chain shall be matched by increased flexibility in the relation with the customer in the shaping of the offer at the retail point.
Hence, it is crucial for retailers to identify emerging needs, potential behaviour changes and trends and drive the commercial strategies accordingly. Constructive Recommendation is a new emerging branch of research in recommendation systems.

Approach

The main innovation associated with this PhD position is the development of a new generation of recommendation systems, making larger use of data a key element. This approach aims at designing a specific recommendation system for the retail world, especially for product or service bundling to create new up-selling and cross-selling opportunities.  Petri Liuha, EIT Digital's Action Line Leader Digital Industry: "The data-driven aspects are of specific importance to new solutions in the Digital Industry action line, and they are very topical at the moment in all segments of retail, regardless of the channel."

The Industrial Doctorates aims at applying these solutions in TIM, for the construction of personalized bundle offers increasing the overall satisfaction of the customer, which may include a wide range of services such as personal or home communication plans, products like mobile devices, smart connected objects and applications such as media content provisioning. The complexity of the problem is increased by the fact that the combination of objects must be controlled by a set of consistency constraints/criteria such as compatibility of products, limits on pricing and must be driven by the trends and opportunities identified by the analysis of the customer's big data provided by TIM.

Expected outcome

This PhD thesis will extend the research line in Constructive Recommendation Systems and deliver innovative recommendation systems prototypes in the domains of interest for TIM, to be applied by TIM in its blended B2C retail channels. The system will be enhanced by the integration of patterns and trends resulting from the analytics of companies' big data sets to drive the evolution of the suggested offers.

Location

The PhD position is based in the Doctoral School Training Centre in Trento. The candidate will spend at least six months in other universities or research centres. For example, three months at the Université Pierre et Marie Curie in Paris to work on algorithmic decision theory approaches for constructive recommendation and three months at the research group of Prof Luc De Raedt at the KU Leuven to explore innovative directions for learning constraints for constructive recommendation.

Facts

  • Industrial partner: Telecom Italia (TIM)
  • Academic/research partner: University of Trento
  • Number of available PhD positions: 1
  • Duration: 3 years 
  • This PhD will be funded by EIT Digital, The University of Trento and TIM.

Apply

Those interested in applying should fill in the application form on the website of The University of Trento.

Please apply before August 31, 2017.

Read here the full position paper

© 2010-2019 EIT Digital IVZW. All rights reserved. Legal notice

Supported by the EIT