The invention discloses a genetic-algorithm-based method for automatically generating data stream test cases. At present, in genetic-algorithm-based technologies of automatically generating data stream test cases, crossover and mutation points of a genetic algorithm are selected at free positions of individual genes randomly, good genes are produced, and meanwhile the possibility that existing good genes are damaged also exists; whether the maximum adaptation degree of each individual in an offspring population is greater than the maximum adaptation degree of each individual in a parent population or not, the offspring population continues to inherit, and an iteration process is prolonged. According to the genetic-algorithm-based method for automatically generating the data stream test cases, a specific branch which does not satisfy a target path is found when an adaptation function is calculated, structural information of the branch is used for constraining a range of crossover operation and mutation operation of the genetic algorithm, and meanwhile offspring with the adaptation degrees being greater than those of parents are abandoned by comparing the maximum adaptation degree of each individual in the offspring population with the maximum adaptation degree of each individual in the parent population; thus, transferring of the good genes is guaranteed, the number of iterations is reduced, and computational efficiency is improved.