Co-evolutionary genetic algorithm in symptom-herb relationship discovery

Traditional Chinese Medicine (TCM) is a holistic approach to medical treatment. The symptoms from a diagnosis are grouped into overlapping sets of symptoms, where each set of symptoms may demand the use of a different set of herbs. Since there are multiple mappings between symptoms and herbs, the discovery of the symptoms-herbs relationship is a crucial step to the research of the underlying TCM principle. The discovery of many existing formulas took a long time to stabilize to the current configurations. In this paper, the relationship discovery is argued to be more than just an evolutionary process, but a co-evolutionary process, i.e. a set of symptoms searches for candidate sets of herbs, while a given set of herbs also searches for multiple sets of symptoms that it can be applied. In other words, a well recognized symptoms-herbs relationship is the result of a dynamic equilibrium of two inter-related evolutionary processes. This model of discovery was implemented using a Combined Gene Genetic Algorithm (CoGA1) where the symptoms and herbs are encoded in the same chromosome to evolve over time. The algorithm was tested with an insomnia dataset from a TCM hospital. The algorithm was able to find the symptoms-herbs relationships that are consistent with TCM principles.