Reinforcement learning method device and program for identifying causal effect in logged data

• Main Author: SEO SUNG WOOK
• Format: Patent
• Language: English; Korean
• Published: 27.08.2021
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATIONTECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING ORPROCESSING OF MEDICAL OR HEALTHCARE DATA; INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTEDFOR SPECIFIC APPLICATION FIELDS; PHYSICS

Provided is a reinforcement learning method for identifying causal effect comprising: a step of acquiring recorded data related to patients; a step of sampling a subject included in another non-treated group from the recorded data to estimate causality of treatment and a result for any target subject; a step of comparing a result of the target subject and a result of the sampled subject to generate causal compensation; and a step of performing reinforcement learning using the causal compensation. 본 발명은, 환자들에 관한 기록된 데이터를 획득하는 단계와, 임의의 대상 개체에 대한 치료와 결과의 인과성을 추정하기 위해, 상기 기록된 데이터로부터, 상기 치료를 받지 않은 상대군에 포함된 개체를 샘플링하는 단계와, 상기 대상 개체의 결과와 상기 샘플링된 개체의 결과를 비교하여, 인과보상을 생성하는 단계와, 상기 인과보상을 사용한 강화학습을 수행하는 단계를 포함하는, 인과성을 식별하는 강화학습 방법을 제공한다.