A generation method executed by a generation device including a processor configured to execute a program and a storage device that stores the program, the generation method comprising:
- executing by the processor:
- visualization element generation processing of generating, based on time series data including feature amounts of a plurality of different factors existing in time series, a feature amount range visualization element that defines a feature amount range, along a vertical axis, that is a continuous range of the feature amount for each of the factors;
- generating an inter-feature amount visualization element that defines relevance between a first feature amount of a first factor and a second feature amount of a second factor that are continuous in time among the plurality of different factors, the first factor and the second factor being different factors used to generate a prediction value of a result;
- and
- visualization information generation processing of generating visualization information indicating a relationship of the feature amounts related to the plurality of different factors by associating, by using the inter-feature amount visualization element, a first feature amount range visualization element of the first factor defined along a first vertical axis with a second feature amount range visualization element of the second factor defined along a second vertical axis among the feature amount range visualization element for each of the factors generated by the visualization element generation processing,
- wherein the first vertical axis and the second vertical axis are displayed in chronological order based on the time series and the inter-feature amount visualization element is displayed as a line segment with one end indicating a first value of the first feature amount along the first vertical axis and another end indicating a second value of the second feature amount,
- wherein the processor is configured to execute analysis processing of generating the prediction value, which is a probability of being a specific result based on inputting the time series data into a prediction model configured to predict that the result related to the plurality of different factors including the first factor and the second factor is the specific result,
- wherein the prediction model is generated by deep learning based on time series data, that includes respective feature amount vectors indicating feature amounts for the plurality of factors, and the feature amount vectors are input to the generated prediction model as test data,
- wherein in the visualization element generation processing, the processor is configured to generate a factor visualization element that defines each of the plurality of different factors and a first relevance degree visualization element that defines a first relevance degree between the first factor and the second factor, and
- wherein, in the analysis processing, the processor is configured to output an importance degree for illustrating the prediction value, and
- wherein, in the visualization element generation processing, the processor is configured to generate the first relevance degree visualization element based on a first importance degree of the first factor and a second importance degree of the second factor output by the analysis processing.