Pattern Structures for Treatment Optimization
A comparison of different treatment strategies does not always result in determining the best one for all patients, one needs to study subgroups of patients with significant difference in efficiency between treatment strategies. To solve this problem an approach to subgroups generation is proposed, where data are described in terms of a pattern structure and pattern concepts stay for patient subgroups and their descriptions. To find the most promising pattern concepts in terms of the difference of treatment strategies in efficiency a version of CbO algorithm is proposed. An application to the analysis of data on childhood acute lymphoblastic leukemia is considered.