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:
- receiving a set of one or more input data objects, wherein an input data object of the set of one or more input data objects comprises a temporal sequence of a plurality of temporal sequences and a subset of a set of one or more features that corresponds to the set of one or more input data objects;
- generating a global guidance correlation graph data object, wherein:
- (i) a node of the global guidance correlation graph data object corresponds to a feature of the set of one or more features, and
- (ii) an edge of the global guidance correlation graph data object (a) corresponds to a feature pair of the set of one or more features, and (b) comprises a weight that corresponds to a co-occurrence probability for the feature pair;
- generating a set of one or more dynamic co-occurrence graph data objects based at least in part on the global guidance correlation graph data object, wherein a dynamic co-occurrence graph data object of the set of one or more dynamic co-occurrence graph data objects (i) corresponds to the temporal sequence and (ii) comprises a projection of the global guidance correlation graph data object on the input data object that corresponds to the temporal sequence;
- generating, using a set of one or more embeddings, a set of one or more historical node representations based at least in part on the set of one or more dynamic co-occurrence graph data objects, wherein the set of one or more embeddings is:
- (i) generated by:
- (a) generating a set of one or more latent representation vectors that correspond to a set of one or more descriptions of the subset of the set of one or more features, and
- (b) generating, based at least in part on the set of one or more latent representation vectors, a set of one or more input vectors that correspond to one or more hierarchical relationships between the subset of the set of one or more features, and
- (ii) modified over a progression through time corresponding to the set of one or more dynamic co-occurrence graph data objects;
- providing the set of one or more historical node representations to a machine learning model to produce a set of one or more predicted classification labels, wherein accuracy of the machine learning model is improved by using the set of one or more embeddings to generate a set of one or more training historical node representations that corresponds to one or more model parameters of the machine learning model.