To overcome the defects of local optimum and precocity in basic fruit fly algorithm, an improved drosophila algorithm was presented in this paper for optimization. The basic idea is to improve the searching ability and accuracy of the algorithm by using the self-non-self antigen recognition mechanism of immune algorithm and the knowledge processing mechanism of learning-memory-forgetting in immune system. The immune response is introduced at the later stage of fruit fly algorithm implementation and the population diversity is enhanced by producing different antibodies to jump out of the local optimum. The results of numerical simulation and practical cases show that the improved algorithm performs better, and it provides an effective and feasible method and idea for algorithm optimization.