The invention discloses a mapreduce computation process optimization method. The method includes the steps of dividing an original data file into a plurality of files, selecting a file from an unprocessed file collection as a sub-job to be input, determining whether files needing to be merged exist or not , and if not, submitting a task; starting a Map task with the same processing process, executing the Map operation, receiving a Map output result after sorting, merging and partitioning the output Map, performing the Reduce operation, and saving the output result; if files needing to be merged exist, submitting a task; starting a Map task with multiple processing, executing the Map operation, sending different input data to the corresponding Map, executing the Map operation, and performing sorting, merging and partitioning for multi-output; finally checking whether unprocessed data files exist in the original data file collection or not , if not, terminating a program, and if so, performing the procedure again on the divided data files. The mapreduce computation process optimization method disperses the output time, decreases the transient network traffic flow, reduces the occupancy rate of a local disk and improves the MapReduce computation process.