Acquisition of Costly Information in Data-Driven Decision Making
| Thesis title in Czech: | Akvizice nákladné informace při rozhodování na základě dat |
|---|---|
| Thesis title in English: | Acquisition of Costly Information in Data-Driven Decision Making |
| Key words: | Nákladná informace, Rozhodování z dat, Strojové učení, Zpětnovazební učení |
| English key words: | Costly information, Data-driven decision-making, Machine learning, Reinforcement learning |
| Academic year of topic announcement: | 2019/2020 |
| Thesis type: | diploma thesis |
| Thesis language: | angličtina |
| Department: | Institute of Economic Studies (23-IES) |
| Supervisor: | doc. PhDr. Jozef Baruník, Ph.D. |
| Author: | hidden - assigned by the advisor |
| Date of registration: | 31.05.2020 |
| Date of assignment: | 31.05.2020 |
| Date and time of defence: | 16.06.2021 09:00 |
| Venue of defence: | Výuka probíhá online, JONLINE, Pomocná místnost pro rozvrhování výuky probíhají online |
| Date of electronic submission: | 03.05.2021 |
| Date of proceeded defence: | 16.06.2021 |
| Opponents: | Mgr. Lukáš Vácha, Ph.D. |
| URKUND check: | |
| References |
| Howard Raiffa and Robert Schlaifer. Applied Statistical Decision Theory. 1st ed. Harvard University Press, 1961. isbn: 0-87584-017-5.
Stuart Russell and Peter Norvig. Artificial Inteligence. A Modern Ap-proach.3rd ed. Prentice Hall Series in Artificial Intelligence. Prentice Hall,2009. isbn:978-0-13-604259-4. Christopher M. Bishop.Pattern Recognition and Machine Learning. SpringerScience+Business Media, 2006. isbn: 978-0387-31073-2. Onur Atan and Mihaela van der Schaar. Data-Driven Online Decision Making with Costly Information Acquisition. 2016. arXiv:1602.03600[stat.ML]. Russell Greiner, Adam J. Grove, and Dan Roth. "Learning Cost-Sensitive Active Classifiers". In:Artif. Intell.139.2 (2002), pp. 137-174.doi:10.1016/S0004-3702(02)00209- 6. Pallika Kanani and Prem Melville. "Prediction-time Active Feature-value Acquisition for Cost-Effective Customer Targeting". In: (Jan. 2008). Kimmo Karkkainen et al. Cost-Sensitive Feature-Value Acquisition Using Feature Relevance. 2019. arXiv:1912.08281 [cs.LG]. Vikas Raykar, Balaji Krishnapuram, and Shipeng Yu. "Designing Efficient Cascaded Classifiers: Tradeoff between Accuracy and Cost". In: June 2010,pp.853-860.doi:10.1145/1835804.1835912. Hajin Shim, Sung Ju Hwang, and Eunho Yang. "Joint Active Feature Acquisition and Classification with Variable-Size Set Encoding". In:NeurIPS.2018. |
| Preliminary scope of work |
| 1. Motivation of the topic, practical importance of studying the topic - applications in medicine, credit scoring, recommendation systems
2. Description of theories and tools relevant to the information optimization problem: • Value of information theory • Tools from machine learning, cost-sensitive machine learning, selection of currently used methods for solving tasks similar to the information optimization problem • Markov decision process, Dynamic programming, Reinforcement learning, selection of currently used methods for solving tasks similar to the information optimization problem 3. Introduction of a simple game demonstrating the information optimalization problem, showing that the game can be extended to a very general form and can serve as a suitable representation of the problem 4. Analysis of the game from the perspective of • Value of information theory • Algorithms on cost-senstive machine learning basis including own suggestion • Reinforcement learning and related approaches 5. Comparison of the methods from a perspective of • Optimality of the solution • Interpretability of the results • Computational complexity 6. Case study - usage of selected approaches on real-world data from a credit scoring 7. Conclusion |
| Preliminary scope of work in English |
| 1. Motivation of the topic, practical importance of studying the topic - applications in medicine, credit scoring, recommendation systems
2. Description of theories and tools relevant to the information optimization problem: • Value of information theory • Tools from machine learning, cost-sensitive machine learning, selection of currently used methods for solving tasks similar to the information optimization problem • Markov decision process, Dynamic programming, Reinforcement learning, selection of currently used methods for solving tasks similar to the information optimization problem 3. Introduction of a simple game demonstrating the information optimalization problem, showing that the game can be extended to a very general form and can serve as a suitable representation of the problem 4. Analysis of the game from the perspective of • Value of information theory • Algorithms on cost-senstive machine learning basis including own suggestion • Reinforcement learning and related approaches 5. Comparison of the methods from a perspective of • Optimality of the solution • Interpretability of the results • Computational complexity 6. Case study - usage of selected approaches on real-world data from a credit scoring 7. Conclusion |