JPWO2020159568A5 - - Google Patents

Claims (14)

A method for implementing machine learning predictive models that use dynamic data selection.
The data prediction comprises the steps of accessing multiple data predictions generated by the trained machine learning model, the data prediction being composed of corresponding observational data, the method further comprising:
A step of calculating the accuracy rate of the machine learning model based on the number of data predictions accessed and the corresponding observation data, and
Includes the step of accessing and repeating the step of calculating using a variable number of data predictions.
The variable number of data predictions are adjusted based on the operations performed during the last iteration.
A method in which training for the machine learning model is triggered if the calculated accuracy rate does not meet the accuracy criteria during a given iteration. The method of claim 1, wherein the variable number of data predictions for the current iteration are adjusted based on whether training was triggered during the previous iteration. The method of claim 2, wherein the previous iteration comprises the iteration immediately preceding the current iteration. The method of claim 2 or 3 , wherein the triggered training comprises retraining or updated training on the trained machine learning model. When the training is triggered, the next iteration of the accessing step and the calculating step uses any of claims 2-4 using the data prediction generated by the machine learning model generated by the triggered training. Or the method described in item 1 . The method according to any one of claims 2 to 5 , wherein the repeating step is executed according to a predetermined time. The method of claim 6, wherein the predetermined time is a predetermined period or a predetermined amount of data prediction having corresponding observation data. If the training was triggered during the previous iteration, the number of said data predictions for the current iteration is increased, and if the training was not triggered during the previous iteration, the number of said data predictions is decreased. The method according to claim 6 or 7 . If the training was triggered during the previous iteration, the number of said data predictions in the current iteration is multiplied by a factor, and if the training was not triggered during the previous iteration, the number of said data predictions is a factor. The method of claim 8, which is divided by. The method according to claim 9, wherein the coefficient is composed of one or more values. 10. The method of claim 10, wherein the value of the coefficient is reduced over a plurality of iterations. A step of ending the accessing step and the calculating step using the variable number of data predictions after performing the plurality of iterations so that the number of set data predictions is determined based on the repeating steps. Further including, after the repeating step, the set number of data predictions is used to calculate the accuracy rate of the machine learning model, the accuracy rate to trigger training of the machine learning model. The method of claim 11 used. A program that causes a computer to perform the method according to any one of claims 1 to 12. The memory in which the program according to claim 13 is stored and
A system comprising a processor that executes the program .