GUEST LECTURE Prof. Filiz Ersöz is a dedicated researcher and educator with a robust background in statistics, operations research, and data science. She earned the Bachelor's degree in Statistics from Anadolu University in 1989, followed by a Master's (1992) and a Doctorate (1998) in Biometrics from Ankara University. Her career began in 1989 as a Statistician at the Turkish Statistical Institute, where she developed a strong foundation in statistical analysis. She later served as a NATO NAAG Force Coordinator and Defense Industry Cooperation Coordinator at the Land Forces Command (2001-2006), applying operational research principles. From 2006 to 2012, she was engaged as a Probability and Stochastic Processes Expert at the Turkish Military Academy. Between 2012 and 2023, she was an academician in the Department of Industrial Engineering at Karabük University, where she taught a variety of courses, including Probability and Statistics, Engineering Statistics, Applied Multivariate Statistical Analysis, Data Analyzing and Machine Learning, Research Methods, Data Science Principles, Data Mining, Simulation and Modeling, Statistical Quality Control, and Decision Analysis. Currently, she is the Head of the Department of Software Engineering at OSTIM Teknik Üniversitesi, where she continue to contribute to the academic community. With over 100 articles published in renowned journals such as ISI, Scopus, and IEEE, and as the author of several books, including Statistics-I, Statistics-II, Data Mining Techniques and Applications, Statistical Data Analysis with IBM SPSS, and Simulation and Modelling, she is committed to advancing knowledge in her field. Her research interests include decision making, statistics, data mining, simulation and modeling, and multi-criteria decision-making techniques. Industrial systems generate large volumes of structured and unstructured data, yet transforming this data into reliable and actionable decisions remains a methodological challenge. This doctoral-level course examines how statistical modeling, machine learning, and artificial intelligence can be systematically integrated into industrial decision processes. Rather than focusing solely on algorithms, the course emphasizes the formulation of real industrial problems within a rigorous analytical framework. It explores how predictive models and advanced analytical techniques support strategic, tactical, and operational decision making under uncertainty. Particular attention is given to model design, validation, robustness, interpretability, and the practical constraints encountered in applied industrial environments. The course adopts a research-oriented perspective and encourages doctoral students to critically evaluate methodological assumptions, compare alternative analytical strategies, and reflect on the evolving role of AI-driven systems in industrial decision contexts. The course aims to: • Develop a structured understanding of data-driven decision processes in industrial settings • Analyze real industrial problems using statistical and machine learning approaches • Strengthen participants’ ability to design research-based analytical models for complex operational environments • Integrate multivariate statistical analysis with machine learning techniques in practical decision contexts • Encourage critical evaluation of AI-driven analytical solutions applied to real-world industrial systems Upon completion of the course, doctoral students will be able to: • Formulate industrial decision problems within a coherent data-driven analytical framework • Translate operational and managerial challenges into structured statistical and machine learning models • Apply multivariate statistical techniques and predictive modeling approaches to high-dimensional industrial datasets • Evaluate model performance in terms of reliability, robustness, generalization, and interpretability • Critically assess methodological limitations in AI-based decision systems • Identify research gaps and propose analytically grounded improvements in industrial decision models

-

The course combines conceptual lectures with analytical discussions and application-oriented case studies derived from real industrial problems. Examples may include financial risk assessment, demand forecasting, quality analytics, operational performance evaluation, and anomaly detection in industrial data. Illustrative analytical implementations (using Python or KNIME where appropriate) will demonstrate how statistical and machine learning models are developed, validated, and interpreted in practical decision-making settings. Emphasis is placed on methodological rigor and research-oriented thinking rather than software-specific training. Participants are encouraged to critically examine modeling assumptions, compare analytical alternatives, and reflect on the theoretical and practical implications of AI-driven decision systems. The course ultimately aims to equip doctoral students with a rigorous analytical perspective for designing and evaluating data-driven decision frameworks in complex industrial environments. Course Plan Lesson 1 – From Industrial Problems to Analytical Structures • Course introduction • Structuring Industrial Decision Problems Strategic, tactical, and operational decision layers Sources of uncertainty in industrial systems Translating managerial problems into analytical formulations Statistical reasoning as the foundation of data-driven decisions Multivariate Analysis and Dimensionality Reduction Characteristics of high-dimensional industrial datasets Principal Component Analysis (PCA) PCA as a feature engineering and dimensionality reduction tool Discriminant analysis in classification contexts Interpretation and limitations of multivariate models Lesson 2 – Machine Learning in Industrial Decision Contexts • Supervised Learning for Predictive Decision Support Regression-based predictive models Classification algorithms for decision support Model training, validation strategies, and overfitting Performance evaluation metrics Illustrative implementations using Python or KNIME • Unsupervised Learning and Pattern Discover Clustering techniques (e.g., k-means, hierarchical clustering) Cluster validation and stability assessment Pattern identification in operational datasets Applied analytical demonstrations using Python or KNIME Lesson 3 – Deep Learning and Advanced Research Perspectives • Deep Learning for Complex Industrial Data Foundations of neural networks Loss functions and backpropagation principles Deep learning architectures for structured and time-series data Regularization, hyperparameter tuning, and generalization challenges Applications in forecasting, anomaly detection, and fault identification - Model Evaluation, Robustness, and Research Directions Cross-validation and comparative model assessment Interpretability and explainability in AI systems Robustness and generalization in real-world industrial data Emerging methodological challenges in AI-driven industrial decision making • Final evaluation

In presenza

Test a risposta multipla - Sviluppo di project work in team

P.D.2-2 - Maggio

20, 21 e 22 maggio, in orario 9.00-13.00, presso il LEP