This study evaluates the accuracy of a machine learning-based algorithm (XGBoost) to predict response to herbal medicine in patients with functional dyspepsia. The algorithm recommends the one with the expected greatest therapeutic benefit out of three common formulations – Yijung-tang (Lizhong-tang), Pyeongwi-san (Pingwei-san) or Shihosogan-tang (Chaihu Shugan-tang). It is a randomized, evaluator-blinded, open-label, multicenter clinical trial with 100 patients from two Korean hospitals. Patients are randomly assigned to the ACCORD (n=50) group, which will receive the algorithm-recommended treatment, or the DISCORD (n=50), which will receive one of the two non-recommended therapies. Treatment lasts 8 weeks, three times a day between meals. Primary outcome is gastrointestinal symptom score, secondary outcomes include dyspepsia score, symptom relief, quality of life and questionnaire results. Other measurements include blood and fecal metabolome, saliva and stool microflora, heart rate variability, tongue, pulse and abdominal diagnosis. The study is registered under KCT0010587.