Inference method generation based on pattern of hypotheses selection (POHS) dimensions in an interactive visual framework
Embodiments are directed to generating an interactive visual framework to review decision making of machine learning model analysis based on employing a machine learning algorithm to identify patterns in scenario data, map those patterns to dimension values, and generate a machine learning model that represents decision sequences leading to specific outcomes. A generated inference method may predict scenario outcomes based on these correlations and a predetermined threshold, providing valuable insight into potential results. An interactive visual framework may be generated based on the dimension values. The interactive visual framework may be dynamically updated to reflect changes in the machine learning model metadata. This interactive visual framework links normalized dimension values, offering intuitive understanding of the underlying decision-making process. Furthermore, the system is capable of generating sequences of inference methods that enhance predictive capabilities and adaptability. This approach allows for flexible scenario analysis and facilitates a deeper understanding of complex systems.
1 . A method comprising:
identifying patterns in scenario data related to one or more scenarios, and mapping the patterns to dimension values, using a machine learning algorithm, wherein the dimension values further comprise: first dimension values mapped to a first set of patterns identified in user input second dimension values mapped to a second set of patterns identified in methods used by the machine learning algorithm to identify patterns in attributes of the scenario data, and third dimension values mapped to a third set of patterns identified in methods used by the machine learning algorithm to connect one or more patterns in the attributes of the scenario data to one or more outcomes of the one or more scenarios;
generating a machine learning model based at least in part on the patterns, wherein the machine learning model is configured to identify correlations and sequences of machine learning model metadata that represent a combination of decision steps leading to an outcome of the one or more scenarios;
generating an interactive visual framework based at least in part on placing nodes in the interactive visual framework, wherein the nodes are generated from the machine learning model metadata for each decision step based on the dimension values, and normalizing and displaying at least a portion of the dimension values as connections between nodes, wherein the interactive visual framework further comprises a chart with a plurality of axes, and wherein the first dimension values represent a first axis, the second dimension values represent a second axis, and the third dimension values represent a third axis; and
in response to changes to the machine learning model metadata, regenerating the interactive visual framework based on the changed machine learning model metadata, comprising:
regenerating the nodes;
renormalizing the dimension values; and
displaying the regenerated nodes based on the renormalized dimension values.
2 . The method of claim 1 , wherein the method further comprises receiving the changes to the machine learning model metadata.
3 . The method of claim 1 , wherein the method further comprises assigning the dimension values to the patterns, using the machine learning algorithm.
4 . The method of claim 1 , wherein the method further comprises receiving the scenario data.
5 . The method of claim 1 , wherein the method further comprises receiving user input describing the scenario data.
6 . The method of claim 1 , wherein the method further comprises generating the connections between nodes based on the third dimension values.
7 . The method of claim 1 , wherein normalized first dimension values explain transformations of the scenario data prior to the machine learning algorithm identifying the patterns, and normalized second dimension values explain transformations applied to the scenario data within the machine learning algorithm.
8 . The method of claim 1 , wherein the method further comprises displaying a connection between nodes in the interactive visual framework, based at least in part on normalized third dimension values, to show a transition of decision steps in arriving at an outcome, wherein normalized third dimension values explain how the machine learning model combined decision steps.
9 . An article of manufacture that comprises:
at least one non-transitory computer readable storage medium configured to retrievably store instructions for one or more processors, wherein the instructions are structured that, when the instructions are executed by the one or more processors, the instructions cause a computer apparatus configured with the one or more processors to perform actions that comprise:
identify patterns in scenario data related to one or more scenarios, and map the patterns to dimension values, based at least in part on a machine learning algorithm, wherein the dimension values further comprise: first dimension values mapped to a first set of patterns identified in user input, second dimension values mapped to a second set of patterns identified in methods used by the machine learning algorithm to identify patterns in attributes of the scenario data, and third dimension values mapped to a third set of patterns identified in methods used by the machine learning algorithm to connect one or more patterns in the attributes of the scenario data to one or more outcomes of the one or more scenarios;
generate a machine learning model based at least in part on the patterns, wherein the machine learning model is configured to identify correlations and sequences of machine learning model metadata that represent a combination of decision steps that led to an outcome of the one or more scenarios;
generate an inference method configured to predict the outcome of the one or more scenarios, based at least in part on the correlations and sequences of machine learning model metadata and a predetermined threshold, and provide access to an outcome prediction determined by the generated inference method as a function of an operational input;
generate an interactive visual framework based at least in part on nodes placed in the interactive visual framework, wherein the nodes are generated from the machine learning model metadata for each decision step based on the dimension values, and normalize and display at least a portion of the dimension values as connections between nodes, wherein the interactive visual framework further comprises a chart with a plurality of axes, and wherein the first dimension values represent a first axis, the second dimension values represent a second axis, and the third dimension values represent a third axis; and
in response to changes to the machine learning model metadata, regenerate the inference method and regenerate the interactive visual framework based on the changed machine learning model metadata, comprising:
provide access to an outcome prediction determined by the regenerated inference method as a function of the operational input;
regenerate the nodes;
renormalize the dimension values; and
display the regenerated nodes based on the renormalized dimension values.
10 . The article of manufacture of claim 9 , wherein further instructions are structured to cause the computer apparatus to perform actions that further comprise generate the inference method from training data.
11 . The article of manufacture of claim 10 , wherein further instructions are structured to cause the computer apparatus to perform actions that further comprise receive the training data.
12 . The article of manufacture of claim 9 , wherein further instructions are structured to cause the computer apparatus to perform actions that further comprise create and deploy a software agent based at least in part on the generated inference method.
13 . The article of manufacture of claim 9 , wherein the dimension values correspond to a configurable aspect of the generated inference method.
14 . The article of manufacture of claim 9 , wherein further instructions are structured to cause the computer apparatus to perform actions that further comprise generate a sequence of inference methods.
15 . The article of manufacture of claim 14 , wherein further instructions are structured to cause the computer apparatus to perform actions that further comprise provide access to an outcome prediction determined by the sequence of generated inference methods.
16 . The article of manufacture of claim 14 , wherein the sequence of generated inference methods comprise an inferencestream.
17 . The article of manufacture of claim 16 , wherein the inferencestream further comprises different inference methods generated by different machine learning algorithms.
18 . The article of manufacture of claim 17 , wherein the different machine learning algorithms are configured to receive different input data.