In order to improve the efficiency of test data generation for path coverage

a method for generating test data was proposed

which was based on automatic division of path and artificial fish-swarm (AFS) algorithm.Firstly

the relations between variables and nodes

and between variables and paths

were analyzed.Based on the analysis an algorithm for automatic division of path was presented

which can automatically judge the impact of variables on sub-paths.Secondly

an improved AFS algorithm was developed based on Levy flying and conjugate gradient.By making use of the result of path division and the improved AFS algorithm

a new method for searching test data was proposed.If there exist sub paths that the fish pass through in the process of using AFS to generate test data

the corresponding component of these fish were fixed

so that search space were reduced.Finally

the proposed method was applied to the test data generation of programs.It is shown that our method outperforms the related methods in running time

success rate and stability.