The invention provides a cloud task scheduling method based on an improved genetic algorithm which is the dual fitness multiplied genetic algorithm, and relates to the fields of cloud computing and scheduling algorithms. As for a cloud computing Map/Reduce programming model, the problem of matching between user tasks and virtual resources is researched to find appropriate resources for execution of the user tasks, the total task execution time and the mean task execution time are shortest at the same time through the algorithm, and according to the algorithm, one fitness and one multiplied algorithm comprehensively adopting task scheduling are added. An initial population is generated, the individual fitness value is calculated, the operations of selection, intersection and mutation are carried out, the number of iteration is added by one, new populations are generated constantly, the individual fitness values of the finally-obtained population are calculated respectively, the individual with the largest fitness value is the optimal solution, and a resource node sequence obtained by decoding the individual is the final result of the cloud task scheduling method based on the improved genetic algorithm.