DISTRIBUTED PROCESSING SYSTEM, LEARNING MODEL CREATING METHOD, DATA PROCESSING METHOD, LEARNING MODEL CREATING PROGRAM, AND DATA PROCESSING PROGRAM

摘要

PROBLEM TO BE SOLVED: To prevent the accuracy of a prediction processing result of a stream processing-based learning model from becoming reduced.SOLUTION: Plural nodes 40 of a distributed processing system 1 applies a learning model to allocated data and processes the data in a predetermined time period with a time stamp of the data used as a reference. The distributed processing system 1 creates an update learning model and sends it to the plural nodes 40. The distributed processing system 1 sends application timing information to the plural nodes 40, the application timing information being data to be applied with the update learning model which corresponds to the update learning model sent to the plural nodes 40. Upon receiving the update learning model and application timing information, the plural nodes 40 apply the update learning model to data corresponding to the timing prior to the application timing information. The plural nodes 40 also apply the update learning model to data corresponding to the timing subsequent to the application timing information.SELECTED DRAWING: Figure 1