Keynotes

Professor Hop Nguyen, Dean, School of Industrial Engineering and Management, HCMC International University, Vietnam National University

Title: Enhancing Digital Transaction Process in E-commerce by Optimization Solutions: Efficient Crowd-Shipping in Last Mile Deliveries

The key challenges in handling E-commerce orders include a high quantity of order transactions and short processing times. The case of crowd-shipping in last-mile deliveries is addressed to illustrate how to handle these challenges with different settings that not only account for multiple factors in route choice but also consider the preference levels of routing options. Popular crowd-shipping systems are investigated to deal with the uncertainty of demand as well as the availability of supply sources. In detail, we must consider not only route profitability but also other factors affecting occasional drivers’ route choices, such as route duration, transshipment nodes, and destination. Additionally, we explore the insights of shared delivery options in the routing process. Advanced techniques are also proposed to provide efficient solutions for these instances, including dynamic game approaches for multiple players embedded with advanced machine learning and optimization algorithms for allocating appropriate orders to the right player and generating efficient routes based on preference levels of delivery options.

Dr Andrej Svetlosak, School of Informatics, Statistical Learning and Financial AI, The University of Edinburgh, UK

Title: A game-inspired interpretable black box algorithm for marginal and global clustering, with application for financial customer segmentation.

Abstract: Reign-and-Conquer clustering is a novel method that takes advantage of the sturdiness of model-based clustering while attempting to mitigate some of its pitfalls. First, we note that standard model-based clustering likely leads to the same number of clusters per margin. This seems a rather artificial assumption for a variety of datasets, such as financial market segmentation. We tackle this issue by specifying a finite mixture model per margin that allows each margin to have a different number of clusters and then cluster the multivariate data using a similarity-based strategy game-inspired algorithm.

The method can be used as an interpretable black box market segmentation algorithm that allocates individuals into groups with little to no need for recalibration. In a financial context, we highlight the algorithm’s usefulness using a dataset containing detailed transactional information of 10,689 customers’ accounts across 70 UK financial institutions.

Professor Huan Nguyen, HCMC University of Economics

Title: Robo-Advisors and Automated Wealth Management

Abstract: Robo-advisors and automated wealth management have emerged as disruptive technologies in the financial industry, revolutionizing the way individuals manage their investments. This paper provides an overview of the concept of robo-advisors and explores their impact on the wealth management landscape. Robo-advisors are digital platforms that utilize algorithms and artificial intelligence to provide automated investment advice and portfolio management services. They offer a user-friendly and cost-effective alternative to traditional human financial advisors, making investment services accessible to a wider audience. The talk delves into the benefits of robo-advisors, including their ability to provide personalized investment strategies based on individual goals, risk tolerance, and time horizon. By leveraging sophisticated algorithms, robo-advisors can analyze vast amounts of data and execute trades efficiently, optimizing portfolio performance.

Dr Victor Medina-Olivares, Research Center Trustworthy Data Science and Security (UA Ruhr) 
Chair of Uncertainty Quantification and Statistical Learning
Department of Statistics, Technische Universität Dortmund

Title: Semi-structured Multi-State Delinquency Model

Abstract: Credit risk analysis is crucial for lenders and regulatory bodies to ensure sound policy-making. Failure to manage credit risk effectively can lead to financial losses, regulatory penalties and even systematic instability. Time-to-event approaches (a.k.a. survival analysis) are widely used in credit risk modelling. They are concerned with predicting if an event will occur and when it will occur. Multi-state models offer a generalization of the traditional survival approaches; however, their use is not widespread in the context of credit risk. We provide a comprehensive multi-state delinquency approach that accounts for structured linear and non-linear effects, as well as unstructured effects, all in a computationally efficient end-to-end deep neural network.