Influencer segmentation detector

1. A method comprising:

selecting transaction data for a time window;

processing the transaction data for the time window to generate a plurality of graphs based on transaction type for the time window, the plurality of graphs comprising a plurality of nodes and a plurality of edges connecting the plurality of nodes;

for each node in the plurality of nodes:

extracting, from the graph plurality of graphs, a feature set for the each node for the time window;

processing the feature set, extracted for the each node for the time window from the transaction data, to generate a predicted rank for the each node for a subsequent time window using a machine learning model and based at least in part on each edge in the plurality of edges connected to the each node; and

selecting, using the predicted rank, an entity identifier corresponding to the node;

selecting the entity identifier by sorting the plurality of nodes based on the predicted ranks for the subsequent time window for the plurality of nodes into a sorted list;

selecting an entity identifier from the sorted list;

transmitting a message using the entity identifier by sending an email with a link; and

presenting the entity identifier.

2. The method of claim 1 ,

wherein content of the message is based on the predicted rank, wherein the entity identifier is part of a list of entities sorted by a plurality of influence levels determined from the predicted rank and a plurality of influence level thresholds.

3. The method of claim 1 , further comprising:

training the machine learning model to generate the predicted rank for the subsequent time window from the feature set for the time window by:

calculating a plurality of calculated ranks, for a plurality of time windows, from a plurality of additional graphs for the plurality of time windows; and

updating the machine learning model using comparisons of a plurality of model outputs to the plurality of calculated ranks.

4. The method of claim 1 , further comprising:

processing the transaction data for a transaction type to generate the plurality of graphs and generate the predicted rank for the transaction type.

5. The method of claim 1 , further comprising:

processing the feature set by the machine learning model, wherein the machine learning model comprises one or more of a recurrent neural network, a long short term memory (LSTM), a gated recurrent unit (GRU), and a transformer neural network.

6. The method of claim 1 , further comprising:

calculating a rank of the each node of the plurality of graphs with a rank algorithm that uses a count of edges of the each node.

7. The method of claim 1 , further comprising:

extracting the feature set comprising one or more attribute features and topological features.

8. A system comprising:

a graph controller configured to generate a graph;

an extraction controller configured to extract a feature set;

a machine learning model configured to generate a predicted rank;

a selection controller configured to select an entity identifier; and

an application executing on one or more servers and configured for:

selecting transaction data for a time window;

processing, by the graph controller, the transaction data for the time window to generate a plurality of graphs based on transaction type for the time window, the plurality of graphs comprising a plurality of nodes and a plurality of edges connecting the plurality of nodes;

for each node in the plurality of nodes:

extracting, by the extraction controller from the plurality of graphs, a feature set for the each node for the time window;

processing, by the machine learning model, the feature set, extracted for the each node for the time window from the transaction data, to generate the predicted rank for the each node for a subsequent time window and based at least in part on each edge in the plurality of edges connected to the each node; and

selecting, by the selection controller using the predicted rank, an entity identifier corresponding to the node;

sorting the plurality of nodes based on the predicted ranks for the subsequent time window for the plurality of nodes into a sorted list;

selecting, by the selection controller, the entity identifier from the sorted list; and

transmitting a message using the entity identifier by sending an email with a link; and

presenting the entity identifier.

9. The system of claim 8 ,

wherein content of the message is based on the predicted rank, wherein the entity identifier is part of a list of entities sorted by a plurality of influence levels determined from the predicted rank and a plurality of influence level thresholds.

10. The system of claim 8 , wherein the application is further configured for:

training the machine learning model to generate the predicted rank for the subsequent time window from the feature set for the time window by:

calculating a plurality of calculated ranks, for a plurality of time windows, from a plurality of additional graphs for the plurality of time windows; and

updating the machine learning model using comparisons of a plurality of model outputs to the plurality of calculated ranks.

11. The system of claim 8 , wherein the application is further configured for:

processing the transaction data for a transaction type to generate the plurality of graphs and generate the predicted rank for the transaction type.

12. The system of claim 8 , wherein the application is further configured for:

processing the feature set by the machine learning model, wherein the machine learning model comprises one or more of a recurrent neural network, a long short term memory (LSTM), a gated recurrent unit (GRU), and a transformer neural network.

13. The system of claim 8 , wherein the application is further configured for:

calculating a rank of the each node of the plurality of graphs with a rank algorithm that uses a count of edges of the each node of the graph.

14. The system of claim 8 , wherein the application is further configured for:

extracting the feature set comprising one or more attribute features and topological features.