Data Science and Machine Learning to Support Business Decisions

Permanent URI for this collectionhttps://hdl.handle.net/10125/112425

Browse

Now showing 1 - 10 of 10

Role-Aware Backbone Extraction and Visualization of Financial Transaction Networks via Asymmetric Non-negative Matrix Factorization
(2026-01-06) Bose, Aparajita; Kim, Byunghoon; Kim, Seungbeom; Choi, Byungchul
This paper proposes a novel backbone extraction framework tailored for financial transaction networks (FTNs), which are inherently directed, weighted, and often dense. Traditional backbone extraction methods typically assume undirected or symmetric structures and struggle to capture the role-specific, directional nature of financial data. To address this issue, we introduce an Asymmetric Non-negative Matrix Factorization (Asymmetric NMF) technique that decomposes FTNs into low-rank representations, preserving directional features while simplifying network complexity. This method effectively isolates the most significant inter-firm financial relationships and identifies influential firms from both buyer and seller perspectives. The model is validated using real-world industrial financial transaction data, demonstrating its superiority over existing backbone extraction methods in interpretability and structural fidelity.
Learning to Balance: Equitable Districting and Routing in Last-Mile Logistics via Graph Neural Networks
(2026-01-06) Haustein, Vanessa; Gust, Gunther
This work provides a data-driven, deep learning-based solution to the districting and routing problem. Related previous solution approaches focus on cost minimization and face limitations by yielding highly imbalanced districts. This imbalance can cause practical problems such as excessive service times, low customer satisfaction, and unfair workload distribution among deliverers. We propose a deep learning-based solution architecture based on Graph Neural Networks that integrates balance-awareness into the learning process. Evaluation on a large set of real-world cities demonstrates that our approach achieves a significant improvement in workload balance.
Health Comes at a Hidden Cost: Strategic Pricing and Equity in Heart Supplement Markets
(2026-01-06) Zeng, Li; Hang, Ziqi
Strategic pricing can improve profitability but may exacerbate disparities among vulnerable consumers. To examine this tension, this work utilizes NielsenIQ data on heart-health supplement purchases, analyzing empirical coupon usage and price disparities associated with household demographics and health status. Based on these models, a simulation study evaluates the impact of four pricing strategies: uniform, second-degree, third-degree, and combined, under varying behavioral assumptions. Findings highlight how revenue gains from demographic-based pricing can come at the cost of increased inequity, but also show that more balanced outcomes which support both profitability and fairness, are achievable through combined or behavior-based strategies.
A Dynamic Capabilities Perspective on Restaurant Demand Forecasting: Insights from a Large Restaurant Chain
(2026-01-06) Chae, Bongsug; Sheu, Chwen; Park, Eunhye
The previous literature on restaurant businesses provides limited insight into demand forecasting. As a theoretical foundation, this research adopts a dynamic capabilities perspective on demand forecasting. From this viewpoint, restaurants possess different levels of demand forecasting capabilities. This study enhances a range of these capabilities across various market scenarios by utilizing different data and analytics tools. By introducing the dynamic capabilities perspective, this research aims to help restaurants assess their demand forecasting capabilities and provide strategic guidance for developing optimal demand forecasting strategies, considering both the technical and organizational resources available within and outside their business, as well as various market conditions.
A Multi-Task Learning Approach for Predicting Capacity Expansion Timing and Requirements in Colocation Datacenters
(2026-01-06) Zarayeneh, Neda; Mavaie, Pegah; Vennelakanti, Ravigopal
Colocation data centers are essential infrastructure for the digital economy, supporting scalable and secure operations across industries. With the global colocation market exceeding $218 billion—fueled by data growth, technology, and economic expansion—intelligent long-term capacity planning is critical. Traditional reactive methods often miss localized demand shifts, leading to under-provisioning or costly over-investment. A key challenge lies in forecasting when and how much capacity will be needed across infrastructure segments, such as power delivery domains. We propose a multi-task learning framework to jointly predict the timing and magnitude of future capacity expansions. Our hybrid Transformer-based architecture integrates static and temporal features, such as facility telemetry, sector metadata, macroeconomic indicators, and sentiment signals, into a unified temporal embedding space with a static feature layer. It generates dual outputs: a binary classifier for expansion events and a conditional regressor for size. By modeling long-range dependencies and uncertainty, our approach enables accurate, adaptive forecasts that support proactive procurement, smarter resource allocation, and improved infrastructure agility
Anomaly Detection in Multivariate Time Series: Combining LSTM Autoencoders with Contrastive Learning
(2026-01-06) Seo, Sooyon; Jang, Sora; Min, Moohong
We propose a Long Short-Term Memory(LSTM)-based autoencoder model for multivariate time series anomaly detection that incorporates contrastive learning tailored to time-series characteristics. By leveraging contrastive representation learning, the model effectively pulls normal data closer to the original representation while pushing anomalous data further away, enhancing detection performance. To generate positive and negative pairs, the model applies time series-specific augmentations by sampling overlapping segments, preserving contextual integrity. It combines both instance and temporal contrastive learning to capture richer representations. Training is guided by a joint loss function that integrates weighted contrastive loss with reconstruction loss. Experimental results demonstrate that the proposed method improves F1-score by 2–6% over baseline models. This work highlights that even with a simple LSTM-based autoencoder architecture, significant gains in anomaly detection can be achieved by incorporating contrastive learning strategies suited for time series data.
Tackling Managerial Cognitive Limitations by Harnessing Artificial Intelligence
(2026-01-06) Niittymaa, Jukka; Ruokonen, Mika
Human cognitive limitations pose significant challenges to processing the vast amounts of data necessary for strategic decision-making and management. This paper examines some of the limitations, including heuristics and biases, and explores how emerging Artificial Intelligence (AI) systems may transform strategic management practices in the future. Drawing on previous literature and industry insights, we trace how AI tools are already reshaping strategic decision-making to overcome human limits. Our paper reveals a tension: while novel AI excels at pattern recognition, explanation, content creation, and predictive analytics across vast datasets, many strategically critical decisions, particularly those involving ethical judgment and authentic relationship building, are likely reserved for humans. This suggests that future leaders must balance AI's analytical power with distinctly human capabilities, and humans benefit from adapting to and collaborating with machines in decision-making processes. The paper concludes by proposing six working hypotheses and delineating research opportunities for understanding human-AI collaboration in the future.
Prompting for the Unknown: Leveraging In-Context-Learning for Few-Shot Open Set Classification
(2026-01-06) Grote, Alexander; Hariharan, Anuja; Weinhardt, Christof
Recognising customer intent is crucial for applications such as chatbots and virtual assistants, requiring accurate interpretation of user inputs. While traditional intent recognition systems depend on large datasets and complex machine learning pipelines, large language models (LLMs) offer competitive performance with significantly less training data through in-context learning (ICL). In this work, we assess the effectiveness of ICL for intent recognition, with a particular focus on detecting out-of-distribution (OOD) inputs. We explore prompting strategies to improve OOD detection and systematically evaluate few-shot classifiers under varying OOD proportions. Our results show that implicit prompting strategies yield better precision for OOD detection, while explicit strategies excel at recall. Moreover, we confirm that LLMs perform comparably to conventional classifiers on in-distribution data. However, a significant fraction of OOD errors are non-overlapping between LLMs and traditional models, highlighting limitations in LLM robustness and suggesting new directions for enhancing generalisation in intent recognition systems.
Optimal Device-Type Inference in Cyber-Physical Systems Using Packet Data and Expert Rules
(2026-01-06) Brown, Nathanael; Gearhart, Jared; Wilson, Amanda; Sahakian, Meghan
As cyber threats to cyber-physical systems continue to escalate, gaining a comprehensive understanding of the system architecture becomes essential for implementing effective defense strategies. However, it is often the case that even system owners lack a complete understanding of their systems. This underscores the need for developing approaches that facilitate the creation of detailed network maps to improve situational awareness and bolster cybersecurity measures. In this paper, we propose a novel approach that integrates mathematical optimization techniques with expert rules to infer device types based on communication patterns observed in passive packet capture (PCAP) data. This methodology serves as a crucial first step in developing detailed network maps, ultimately strengthening the security posture of cyber-physical systems.
Introduction to the Minitrack on Data Science and Machine Learning to Support Business Decisions
(2026-01-06) Delen, Dursun; Davazdahemami, Behrooz; Zolbanin, Hamed

Browse

Recent Submissions