Evolutionary Inspired Adaptation of Exercise Plans for Increasing Solution Variety

An initial case base population naturally lacks diversity of solutions. In order to overcome this cold-start problem, we present how genetic algorithms (GA) can be applied. The work presented in this paper is part of the selfBACK EU project and describes a case-based recommendation system that creates exercise plans for patients with non-specific low back pain (LBP). In selfBACK Case-Based Reasoning (CBR) is used as its main methodology for generating patient-specific advice for managing non-specific LBP. The sub-module of selfBACK presented in this work focuses on the adaptation process of exercise plans: A GA inspired method is created to increase the variation of personalized exercise plans, which today are crafted by medical professionals. Experiments are conducted using real patients’ characteristics with expertcrafted solutions and automatically generated solutions. In the evaluation we compare the quality of the GA-generated solutions to null-adaptation solutions.