Interactable and interpretable temporal disease risk profiles
Various embodiments provide methods, apparatus, systems, computing entities, and/or the like, providing a temporal disease risk profile describing a likelihood of disease onset over time for an individual in a dynamically interpretable manner. Interpretability of the temporal disease risk profile is enabled by providing additional and contextual information, such as weight distributions of various health indicators, factors, and features. In an embodiment, an example method comprises generating a temporal disease risk profile comprising risk score nodes based at least in part on providing a plurality of record data objects to a risk scoring machine learning model configured to generate a risk score; providing the temporal disease risk profile for display via a first user interface comprising a plurality of interactable node mechanisms each corresponding to a risk score node; and providing a node-specific weight distribution comprising one or more sub-nodal weight values for display via a second user interface.
1 . A computer-implemented method comprising:
providing for display, by one or more processors and via a first interactable user interface, a profile associated with a machine learned model, wherein:
(i) the first interactable user interface comprises (a) an interactive graph that comprises a horizontal axis defined by a plurality of time bins and a vertical axis defined by a plurality of model outputs from the machine learned model, (b) a plurality of interactable node mechanisms respectively corresponding to a plurality of nodes of the profile, (c) a plurality of nodal weight indications respectively corresponding to the plurality of nodes, and (d) a multi-level interpretability mechanism configured to display, upon selection, a multi-level weight table of the profile,
(ii) a node of the plurality of nodes is displayed within the interactive graph (a) at a horizontal position along the horizontal axis based at least in part on one of the plurality of time bins corresponding to the node and (b) at a vertical position of the vertical axis based on a one of the plurality of model outputs corresponding to the node,
(iii) an interactable node mechanism of the plurality of interactable node mechanisms that corresponds to the node is overlaid at the horizontal position and the vertical position, and
responsive to a selection of the interactable node mechanism that corresponds to the node via the first interactable user interface, providing for display, by the one or more processors and via a second interactable user interface, one or more feature weights and one or more sub-nodal features of the node, wherein the one or more feature weights and the one or more sub-nodal features of the node are retrieved from a data store corresponding to the machine learned model; and
responsive to a selection of the multi-level interpretability mechanism, dynamically providing for display, by the one or more processors and via the first interactable user interface, the multi-level weight table configured to, for the node, provide a contextual description for (i) the one or more sub-nodal features, and (ii) the one or more feature weights.
2 . The computer-implemented method of claim 1 , wherein a feature weight of the one or more feature weights is one of a node-specific weight distribution.
3 . The computer-implemented method of claim 1 , wherein the second interactable user interface indicates a percent change from a preceding model output of a preceding node to the model output of the node.
4 . The computer-implemented method of claim 1 , further comprising providing a multi-disease risk table for display, the multi-disease risk table indicating one or more multi-disease risk scores, wherein the machine learning model is configured to generate a multi-disease risk score.
5 . The computer-implemented method of claim 1 , wherein the multi-level weight table comprises: (i) a code column that identifies a sub-nodal feature of the one or more sub-nodal features of the node, (ii) a code description column that describes the sub-nodal feature, (iii) a code importance column that describes one or more corresponding second feature weights of the sub-nodal feature, and (iv) a visit importance column that describes the first feature weight of the node.
6 . A system comprising:
one or more processors; and
one or more memories storing processor-executable instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:
providing for display, via a first interactable user interface, a profile associated with a machine learned model, wherein:
(i) the first interactable user interface comprises (a) an interactive graph that comprises a horizontal axis defined by a plurality of time bins and a vertical axis defined by a plurality of model outputs from the machine learned model, (b) a plurality of interactable node mechanisms respectively corresponding to a plurality of nodes of the profile, (c) a plurality of nodal weight indications respectively corresponding to the plurality of nodes, and (d) a multi-level interpretability mechanism configured to display, upon selection, a multi-level weight table of the profile,
(ii) a node of the plurality of nodes is displayed within the interactive graph (a) at a horizontal position along the horizontal axis based at least in part on one of the plurality of time bins corresponding to the node and (b) at a vertical position of the vertical axis based on one of the plurality of model outputs corresponding to the node,
(iii) an interactable node mechanism of the plurality of interactable node mechanisms that corresponds to the node is overlaid at the horizontal position and the vertical position, and
responsive to a selection of the interactable node mechanism that corresponds to the node via the first interactable user interface, providing for display, via a second interactable user interface, one or more feature weights and one or more sub-nodal features of the node, wherein the one or more feature weights and the one or more sub-nodal features of the node are retrieved from a data store corresponding to the machine learned model; and
responsive to a selection of the multi-level interpretability mechanism, dynamically providing for display, via the first interactable user interface, the multi-level weight table configured to, for the node, provide a contextual description for (i) the one or more sub-nodal features, and (ii) the one or more feature weights.
7 . The system of claim 6 , wherein a feature weight of the one or more feature weights is one of a node-specific weight distribution.
8 . The system of claim 6 , wherein the second interactable user interface indicates a percent change from a preceding model output of a preceding node to the model output of the node.
9 . The system of claim 6 , further comprising providing a multi-disease risk table for display, the multi-disease risk table indicating one or more multi-disease risk scores, wherein the machine learning model is configured to generate a multi-disease risk score.
10 . One or more non-transitory computer-readable media storing processor-executable instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:
providing for display, via a first interactable user interface, a profile associated with a machine learned model, wherein:
(i) the first interactable user interface comprises (a) an interactive graph that comprises a horizontal axis defined by a plurality of time bins and a vertical axis defined by a plurality of model outputs from the machine learned model, (b) a plurality of interactable node mechanisms respectively corresponding to a plurality of nodes of the profile, (c) a plurality of nodal weight indications respectively corresponding to the plurality of nodes, and (d) a multi-level interpretability mechanism configured to display, upon selection, a multi-level weight table of the profile,
(ii) a node of the plurality of nodes is displayed within the interactive graph (a) at a horizontal position along the horizontal axis based at least in part on a time bin corresponding to the node and (b) at a vertical position of the vertical axis based on a model output corresponding to the node,
(iii) an interactable node mechanism of the plurality of interactable node mechanisms that corresponds to the node is overlaid at the horizontal position and the vertical position, and
responsive to a selection of the interactable node mechanism that corresponds to the node via the first interactable user interface, providing for display, via a second interactable user interface, one or more feature weights and one or more sub-nodal features of the node, wherein the one or more feature weights and the one or more sub-nodal features of the node are retrieved from a data store corresponding to the machine learned model; and
responsive to a selection of the multi-level interpretability mechanism, dynamically providing for display, via the first interactable user interface, the multi-level weight table configured to, for the node, provide a contextual description for (i) the one or more sub-nodal features, and (ii) the one or more feature weights.
11 . The one or more non-transitory computer-readable media of claim 10 , wherein a feature weight of the one or more feature weights is one of a node-specific weight distribution.
12 . The one or more non-transitory computer-readable media of claim 10 , wherein the second interactable user interface indicates a percent change from a preceding model output of a preceding node to the model output of the node.