Graduation Semester and Year
Spring 2026
Language
English
Document Type
Dissertation
Degree Name
Doctor of Philosophy in Industrial Engineering
Department
Industrial and Manufacturing Systems Engineering
First Advisor
Shouyi Wang
Second Advisor
Mohammad Jahanbakht
Third Advisor
Paul J. Componation
Fourth Advisor
Habeeb Olufowobi
Abstract
Artificial intelligence is transforming enterprise decision-making, enabling systems that were previously rule-based or manual to become adaptive and data-driven. This dissertation examines how AI integrates into two enterprise decision domains, dynamic pricing and customer relationship management, through five studies contributing synthesis, methodology, deployment, and empirical evidence.
The first three studies address AI-based dynamic pricing. A systematic bibliometric analysis of 1,301 Scopus-indexed publications identifies nine research clusters and documents a fieldwide shift from rule-based and econometric approaches toward deep learning and reinforcement learning, with data-intensive sectors advancing faster than traditional revenue management domains. The dissertation introduces STime-Net, a sparse deep neural network for multivariate time series forecasting that applies hierarchical Group LASSO regularization for simultaneous variable-level and lag-level feature selection within a single training procedure. On synthetic and real-world benchmarks, STime-Net achieves competitive forecast accuracy alongside higher sparsity and stronger noise rejection than regularized baselines, and produces interpretable input records as a direct output of training. STime-Net is embedded as the demand oracle within a predict-thenoptimize framework for e-commerce dynamic pricing; over a 358-day simulation, the adaptive policy achieves $519K in cumulative profit, exceeding a perfect-information greedy oracle and all static benchmarks while reducing profit volatility.
The demonstrated value of AI integration in pricing raises a broader question: how does AI perform when integrated into other enterprise systems? The final two studies address this through customer relationship management. A bibliometric and systematic review of 810 publications maps AI-based CRM into four research clusters, identifying strategic AI implementation as consistently underrepresented. A mixed-methods empirical study grounded in the Resource-Based View finds that business architecture is the significant predictor of effective AI-CRM integration — not human capital or IT infrastructure, both externalized by SaaS platforms — and that AI adoption produces measurable improvements across process, customer, and infrastructure performance dimensions.
Together, these five studies contribute a unified theoretical and empirical foundation for understanding how AI transforms enterprise systems, while providing practitioners with actionable frameworks for deployment in dynamic pricing and CRM contexts.
Keywords
Dynamic Pricing, Sparsity, Feature Selection, Artificial Intelligence, Neural Networks, Multivariate Time Series, Revenue Management, Demand Forecasting, Enterprise Information System
Disciplines
Industrial Engineering | Operational Research
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Recommended Citation
Ozay, Dervis, "STRUCTURED SPARSE DEEP NEURAL NETWORKS FOR MULTIVARIATE TIME SERIES FORECASTING: FROM INTERPRETABLE FEATURE SELECTION TO ENTERPRISE DECISION SUPPORT" (2026). Industrial, Manufacturing, and Systems Engineering Dissertations. 2.
https://mavmatrix.uta.edu/industrialmanusys_dissertations2/2
Comments
Two chapters of this dissertation have been previously published in scholarly journals. In both cases, I retained — under the standard publishing agreement — the right to include the article in my dissertation and to have it posted to my university's institutional repository. No separate written permission is required by either publisher for this use, as documented in their publicly stated author-rights policies. Details for each chapter are below.
Chapter 6 — Published in Electronic Commerce Research and Applications (Elsevier)
Citation: Ozay, D., Jahanbakht, M., & Wang, S. (2025). What resources are needed for effective AI implementation in CRM, and does it actually enhance performance? Electronic Commerce Research and Applications, 74, 101552.
Author rights retained under the Elsevier Journal Publishing Agreement explicitly include the right to include the article, in full or in part, in a thesis or dissertation for non-commercial purposes, with proper acknowledgment. This right extends to posting the thesis in the author's institutional repository, provided the article is embedded in the thesis and not made separately downloadable. No written permission from Elsevier is required. See Elsevier's official policy: https://www.elsevier.com/about/policies-and-standards/copyright/permissions (section: "Can I include/use my article in my thesis/dissertation?").
Chapter 5 — Published in Enterprise Information Systems (Taylor & Francis / Routledge)
Citation: Ozay, D., Jahanbakht, M., Shoomal, A., & Wang, S. (2024). Artificial Intelligence (AI)-based Customer Relationship Management (CRM): a comprehensive bibliometric and systematic literature review with outlook on future research. Enterprise Information Systems, 18(7), Article 2351869. https://doi.org/10.1080/17517575.2024.2351869
Author rights retained under the Taylor & Francis publishing agreement explicitly include the right to include the article in a thesis or dissertation, and to post the Author's Original Manuscript (AOM) / Accepted Manuscript (AM) in an institutional repository (subject to the journal's embargo period, where applicable). The version of this article embedded in the dissertation is the [Accepted Manuscript / Version of Record — choose one], in keeping with this policy. See Taylor & Francis Author Services: https://authorservices.taylorandfrancis.com/publishing-your-research/moving-through-production/copyright-for-journal-authors/.
Both chapters are properly acknowledged in the dissertation, including a chapter title-page footnote and full citation, in accordance with UTA's Article-Based Thesis/Dissertation guidelines.