This study describes a method that uses a genetic algorithm to select optimal beam angles in proton therapy and evaluates the effectiveness of the proposed algorithm in actual patients. In the use of the genetic algorithm to select the optimal angle, a gene represents the angle of each field and a chromosome represents the combination of beam angles. The fitness of the genetic algorithm, which represents the suitability of the chromosome to the solution, was quantified by using the dose distribution. The weighting factors of the organs used for fitness were obtained from clinical data through logistic regression, reflecting the dose characteristics of actual patients. Genetic operations, such as selection, crossover, mutation, and replacement, were used to modify the population and were repeated until an evaluation based on fitness reached the termination criterion. The proposed genetic algorithm was tested by assessing its ability to select optimal beam angles in three patients with liver cancer. The optimal results for fitness, planning target volume (PTV), normal liver, and skin in the population were compared with the clinical treatment plans, a process that took an average of 36.8 minutes. The dose-volume histograms (DVHs) and the fitness of the genetic algorithm plans did not differ significantly from the actual treatment plans. These findings indicate that the proposed genetic algorithm can automatically generate proton treatment plans with the same quality as actual clinical treatment plans.