The invention provides a workflow scheduling method for data analysis tasks, and relates to the field of computer software. The method includes the steps that S1, a task sequence is received, and minimum task units in the task sequence are parsed and packaged; S2, a data flow diagram is constructed according to the minimum task units, wherein the data flow diagram includes nodes; S3, the nodes inthe data flow diagram are initialized; S4, the initialized data flow diagram is searched for parallel nodes; S5, tasks are executed through multiple worker threads according to the data flow diagram,wherein the parallel nodes execute tasks simultaneously; S6, steps S4-S5 are circularly conducted until execution of the task with an ending tag is finished, and then the whole workflow scheduling iscompleted. The method solves the problem that in most current scheduling methods, one node manager takes charge of task execution of all nodes, and once one node program runs incorrectly or the node manager breaks down, execution of the whole data analysis task fails.