Visualization device, visualization method and visualization program
In accordance with a score representing a relationship between features of input data, an analysis unit analyzes the importance of each feature. An extraction unit extracts a combination of features in which the score satisfies a predetermined condition. A specification unit specifies a visualization system for the combination in accordance with at least the types and connection relationships of the features constituting the extracted combination. A drawing unit draws the relationship between the features constituting the combination in order of priority according to the importance, based on the combination of the features and the visualization system corresponding to the combination.
1 . A visualization device, comprising:
a memory; and
a processor coupled to the memory and programmed to perform operations comprising:
obtaining input data;
extracting a plurality of features from the input data;
identifying pairs of features from the plurality of features;
calculating weights respectively corresponding to the pairs of features based on at least one of a correlation analysis, an importance analysis, a prediction rule analysis, or a responsiveness analysis, wherein the respective weight indicates a strength of a relationship between the corresponding pair of features;
generating, using the calculated weights, an adjacency matrix of the pairs of features;
calculating centrality values of the plurality of features based on a centrality analysis of the adjacency matrix of the pairs of features;
extracting a combination of features from the plurality of features based on the centrality values of the plurality of features;
identifying a visualization system for the extracted combination of features in accordance with at least types and connection relationships of features included in the extracted combination of features; and
drawing a relationship between the features included in the extracted combination of features in order of priority according to the calculated centrality values, based on the extracted combination of features and the visualization system corresponding to the extracted combination of features.
2 . The visualization device according to claim 1 , wherein calculating the weights respectively corresponding to the pairs of features comprises training a machine learning model in which one of the plurality of features is constructed as an objective variable and at least one other feature of the plurality of features is constructed as an explanatory variable.
3 . The visualization device according to claim 2 , wherein training the machine learning model comprises approximating a prediction result of the trained machine learning model using a decision tree.
4 . A visualization device, comprising:
a memory; and
a processor coupled to the memory and programmed to perform operations comprising:
presenting centrality values of a plurality of features, wherein:
the centrality values of the plurality of features are calculated based on a centrality analysis of an adjacency matrix of pairs of features;
the pairs of features are identified from the plurality of features;
the adjacency matrix is generated using weights respectively corresponding to the pairs of features; and
the weights are calculated based on at least one of a correlation analysis, an importance analysis, a prediction rule analysis, or a responsiveness analysis, wherein the respective weights indicate a strength of a relationship between the corresponding pair of features;
obtaining a feature selected by a user based on the presented centrality values of the plurality of features; and
drawing a relationship between features including the combination of features including the selected feature, based on the combination of features including the selected feature and a predetermined visualization system corresponding to the combination of features including the selected feature.
5 . A visualization method executed by a visualization device, comprising:
obtaining input data;
extracting a plurality of features from the input data;
identifying pairs of features from the plurality of features;
calculating weights respectively corresponding to the pairs of features based on at least one of a correlation analysis, an importance analysis, a prediction rule analysis, or a responsiveness analysis, wherein the respective weight indicates a strength of a relationship between the corresponding pair of features;
generating, using the calculated weights, an adjacency matrix of the pairs of features;
calculating centrality values of the plurality of features based on a centrality analysis of the adjacency matrix of the pairs of features;
extracting a combination of features from the plurality of features based on the centrality values of the plurality of features;
identifying a visualization system for the extracted combination of features in accordance with at least types and connection relationships of features included in the extracted combination of features; and
drawing a relationship between the features included in the extracted combination of features in order of priority according to the calculated centrality values, based on the extracted combination of features and the visualization system corresponding to the extracted combination of features.
6 . The visualization method according to claim 5 , wherein calculating the weights respectively corresponding to the pairs of features comprises training a machine learning model in which one of the plurality of features is constructed as an objective variable and at least one other feature of the plurality of features is constructed as an explanatory variable.
7 . The visualization method according to claim 6 , wherein training the machine learning model comprises approximating a prediction result of the trained machine learning model using a decision tree.
8 . A non-transitory computer-readable recording medium having stored a visualization program that, when executed by a processor, causes the processor to perform operations, comprising:
obtaining input data;
extracting a plurality of features from the input data;
identifying pairs of features from the plurality of features;
calculating weights respectively corresponding to the pairs of features based on at least one of a correlation analysis, an importance analysis, a prediction rule analysis, or a responsiveness analysis, wherein the respective weight indicates a strength of a relationship between the corresponding pair of features;
generating, using the calculated weights, an adjacency matrix of the pairs of features;
calculating centrality values of the plurality of feature based on a centrality analysis of the adjacency matrix of the pairs of features;
extracting a combination of features from the plurality of features based on the centrality values of the plurality of features;
identifying a visualization system for the extracted combination of features in accordance with at least types and connection relationships of features included in the extracted combination of features; and
a drawing step of drawing a relationship between the features included in the extracted combination of features in order of priority according to the calculated centrality values, based on the extracted combination of features and the visualization system corresponding to the extracted combination of features.
9 . The non-transitory computer-readable medium according to claim 8 , wherein calculating the weights respectively corresponding to the pairs of features comprises training a machine learning model in which one of the plurality of features is constructed as an objective variable and at least one other feature of the plurality of features is constructed as an explanatory variable.
10 . The non-transitory computer-readable medium according to claim 9 , wherein training the machine learning model comprises approximating a prediction result of the trained machine learning model using a decision tree.