Machine learning system for generating recommended electronic actions

1 . A computer implemented system comprising:

an electronic data store comprising a plurality of data records, each data record comprising at least two different types of input data for an entity, wherein the at least two different types of input data comprise categorical variables and free-form textual data and associated properties received as graphical user interface (GUI) input fields on user interface elements of a GUI of a requesting device;

one or more hardware processors in communication with a computer readable medium storing software instructions that are executable by the one or more hardware processors in order to cause the computer implemented system to:

direct a first type of input data corresponding to the free-form textual data to a pre-trained natural language processor and deriving therefrom categorical and continuous attributes via the pre-trained natural language processor, wherein the pre-trained natural language processor comprises one or more pre-trained language models that applies at least one of an Embedding from Language Model (ELMO) or Bidirectional Representations from Transformers (BERT) to derive the categorical and continuous attributes;

feed a defined set of demographical data to the pre-trained natural language processor and combine an output of the derived categorical and continuous attributes with the defined set of demographic data to a candidate data store;

direct a second type of input data corresponding to the categorical variables to a graph processor to generate an entity graph also configured to receive an input of other prior data records and corresponding attributes to determine relationships in the entity graph and a comparison between common attributes of current and prior data records;

feed the entity graph to a graph convolutional network and determining determine common digital paths for the prior data records to reach a given state providing an entity graph output via the graph convolutional network; and

combine the entity graph output and information retrieved from the candidate data store using a machine learning ensemble model to provide a single output prediction of intelligent query responses comprising recommended visualization of recommended digital actions for the requesting device, wherein the machine learning ensemble model applies one of random forest modelling and decision tree modelling to combine results of outputs from processing each of the first and second types of input data via the pre-trained natural language processor and graph convolutional network respectively.

2 . The system of claim 1 , wherein the software instructions are executable by the one or more hardware processors to further cause the computer implemented system to: in response to receiving a user input selecting the recommended digital action, presenting a digital resource for performing the recommended digital action on a display screen of the requesting device.

3 . The system of claim 1 , wherein the graph convolutional network receives as input both an end desired state of the entity, wherein the end desired state is associated with a current data record and derived from the input data received on the user interface elements of the GUI and prior data records of other entities currently at a given state matching the end desired state.

4 . The system of claim 1 , wherein the candidate data store stores profiles of a plurality of candidates and associated attributes.

5 . The system of claim 1 , wherein the software instructions executable by the one or more hardware processors are further configured to cause the computer implemented system to: utilize the pre-trained natural language processor to categorize the free-form textual data into categorical and continuous attributes.

6 . The system of claim 1 , wherein the one or more pre-trained language models provide context to textual inputs received by applying surrounding text in a given sentence of the free-form textual data to establish said context.

7 . The system of claim 3 , wherein the instructions executable by the one or more hardware processors are further configured to cause the computer implemented system to: generate the single output as a set of selectable GUI elements providing access to computer resources for modifying attributes associated with a current state of the entity to achieve the end desired state of the entity.

8 . The system of claim 1 , wherein the instructions executable by the one or more hardware processors are further configured to: receive feedback input to modify categorization of attributes into the categorical and continuous attributes as provided by the pre-trained natural language processor, thereby refining classifications of categories of attributes from the pre-trained natural language processor for subsequent iterations based on the feedback input.

9 . A computer implemented method comprising:

capturing, via a machine learning engine associated with a processor, a plurality of data records, each data record comprising at least two different types of input data for an entity, wherein the at least two different types of input data comprise categorical variables and free-form textual data and associated properties received as graphical user interface (GUI) input fields on user interface elements of a GUI of a requesting device;

directing, via the machine learning engine, a first type of input data corresponding to the free-form textual data to a pre-trained natural language processor and deriving therefrom categorical and continuous attributes via the pre-trained natural language processor, wherein the pre-trained natural language processor comprises one or more pre-trained language models that applies at least one of an Embedding from Language Model (ELMO) or Bidirectional Representations from Transformers (BERT) to derive the categorical and continuous attributes;

feeding, via the machine learning engine, a defined set of demographical data to the pre-trained natural language processor and combining an output of the derived categorical and continuous attributes with the defined set of demographic data to a candidate data store;

directing, a second type of input data corresponding to the categorical variables to a graph processor, to generate an entity graph also configured to receive an input of other prior data records and corresponding attributes to determine relationships in the entity graph and a comparison between common attributes of current and prior data records;

feeding the entity graph to a graph convolutional network and determining common digital paths for the prior data records to reach a given state providing an entity graph output via the graph convolutional network; and

combining, using a machine learning ensemble model, the entity graph output and information retrieved from the candidate data store to provide a single output prediction of intelligent query responses comprising recommended visualization of recommended digital actions for the requesting device, wherein the machine learning ensemble model applies one of random forest modelling and decision tree modelling to combine results of outputs from processing each of the first and second types of input data via the pre-trained natural language processor and graph convolutional network respectively.

10 . The computer implemented method of claim 9 , wherein in response to receiving a user input selecting the recommended digital action, the method comprises presenting a digital resource for performing the recommended digital action on a display screen of the requesting device.

11 . The computer implemented method of claim 9 , wherein the graph convolutional network receives as input both an end desired state of the entity, wherein the end desired state is associated with a current data record and derived from the input data received on the user interface elements of the GUI and prior data records of other entities currently at a given state matching the end desired state.

12 . The computer implemented method of claim 9 , wherein the candidate data store stores profiles of a plurality of candidates and associated attributes.

13 . The computer implemented method of claim 9 , further comprising utilizing the pre-trained natural language processor to categorize the free-form textual data into categorical and continuous attributes.

14 . The computer implemented method of claim 9 , wherein the one or more pre-trained language models provide context to textual inputs received by applying surrounding text in a given sentence of the free-form textual data to establish said context.

15 . The computer implemented method of claim 11 , further comprising generating the single output as a set of selectable GUI elements providing access to computer resources for modifying attributes associated with a current state of the entity to achieve the end desired state of the entity.

16 . The computer implemented method of claim 9 further comprising receiving feedback input on a second user interface associated with the machine learning engine to modify categorization of attributes into the categorical and continuous attributes as provided by the pre-trained natural language processor, thereby refining classifications of categories of attributes from the pre-trained natural language processor for subsequent iterations based on the feedback input.

17 . A non-transitory computer readable medium having instructions tangibly stored thereon, wherein the instructions, when executed cause a computerized system to:

capture, via a machine learning engine associated with a processor, a plurality of data records, each data record comprising at least two different types of input data for an entity, wherein the at least two different types of input data comprise categorical variables and free-form textual data and associated properties received as graphical user interface (GUI) input fields on user interface elements of a GUI of a requesting device;

direct, via the machine learning engine, a first type of input data corresponding to the free-form textual data to a pre-trained natural language processor and deriving therefrom categorical and continuous attributes via the pre-trained natural language processor, wherein the pre-trained natural language processor comprises one or more pre-trained language models that applies at least one of an Embedding from Language Model (ELMO) or Bidirectional Representations from Transformers (BERT) to derive the categorical and continuous attributes;

feed, via the machine learning engine, a defined set of demographical data to the pre-trained natural language processor and combine an output of the derived categorical and continuous attributes with the defined set of demographic data to a candidate data store;

direct, a second type of input data corresponding to the categorical variables to a graph processor, to generate an entity graph also configured to receive an input of other prior data records and corresponding attributes to determine relationships in the entity graph and a comparison between common attributes of current and prior data records;

feed the entity graph to a graph convolutional network and determine common digital paths for the prior data records to reach a given state providing an entity graph output via the graph convolutional network; and

combine, using a machine learning ensemble model, the entity graph output and information retrieved from the candidate data store to provide a single output prediction of intelligent query responses comprising recommended visualization of recommended digital actions for the requesting device, wherein the machine learning ensemble model applies one of random forest modelling and decision tree modelling to combine results of outputs from processing each of the first and second types of input data via the pre-trained natural language processor and graph convolutional network respectively.