Next Best Action: Review and Case Study

Liudmila Kalashnikova

Master's thesis

Abstract:

The development of information technologies and data analysis in particular transformed significantly the processes of decision-making. Companies and institutions strive to make data-driven decisions, and this interest continues to grow. As a result, the field of recommender systems requires new research; therefore, the study of the Next Best Action approach and the features of its application in practice is a task of current interest. The main purpose of the work is to explore the current state of the Next Best Action approach in the data science community by the example of the Covid-19 Government Response Tracker. In the first theoretical part of the work, a review of NBA concept and an analysis of publications on the topic are presented. The chapter includes a justification of the relevance of the topic. In the second part of the work, data description for case study, primary data analysis, application of Next Best Action via Action Rules and Reinforcement Learning and evaluation of the results are proposed. The third part of the work covers development of the Flask-based web-interface. The final outcome of the work is the interactive web-application that allows determining the next best action for preventing Covid-19, the outcome of the interface is based on the initial parameters set by a user and it uses the NBA models, which correspond to current trends in the field. The obtained results of the work could be applied to optimize their activities by data scientists; moreover, case study results might potentially provide a useful tool for government institutions to exchange experience in fighting Covid-19 between countries and support decision-making in this area.

Abstract:

The development of information technologies and data analysis in particular transformed significantly the processes of decision-making. Companies and institutions strive to make data-driven decisions, and this interest continues to grow. As a result, the field of recommender systems requires new research; therefore, the study of the Next Best Action approach and the features of its application in practice is a task of current interest. The main purpose of the work is to explore the current state of the Next Best Action approach in the data science community by the example of the Covid-19 Government Response Tracker. In the first theoretical part of the work, a review of NBA concept and an analysis of publications on the topic are presented. The chapter includes a justification of the relevance of the topic. In the second part of the work, data description for case study, primary data analysis, application of Next Best Action via Action Rules and Reinforcement Learning and evaluation of the results are proposed. The third part of the work covers development of the Flask-based web-interface. The final outcome of the work is the interactive web-application that allows determining the next best action for preventing Covid-19, the outcome of the interface is based on the initial parameters set by a user and it uses the NBA models, which correspond to current trends in the field. The obtained results of the work could be applied to optimize their activities by data scientists; moreover, case study results might potentially provide a useful tool for government institutions to exchange experience in fighting Covid-19 between countries and support decision-making in this area.

Keywords

Machine Learning, Next Best Action, Action Rules, Reinforcement Learning

Keywords

Next Best Action, Machine Learning, Action Rules, Reinforcement Learning

Language used: English

Date on which the thesis was submitted / produced: 30. 6. 2022

Thesis defence

• Date of defence: 25. 8. 2022
• Supervisor: Tomáš Kliegr
• Reader: Lukáš Sýkora

Vysoká škola ekonomická v Praze

Master programme:
Information Systems Management