US 12,248,956 B2
Utilizing machine learning models for data-driven customer segmentation
Luca Costabello, Newbridge (IE); Sumit Pai, Dublin (IE); Fiona Brennan, Dublin (IE); and Adrianna Janik, Dublin (IE)
Assigned to Accenture Global Solutions Limited, Dublin (IE)
Filed by Accenture Global Solutions Limited, Dublin (IE)
Filed on Sep. 30, 2021, as Appl. No. 17/449,506.
Claims priority of provisional application 63/199,234, filed on Dec. 15, 2020.
Prior Publication US 2022/0188850 A1, Jun. 16, 2022
Int. Cl. G06Q 30/0204 (2023.01); G06N 3/042 (2023.01); G06N 5/02 (2023.01); G06Q 30/0251 (2023.01)
CPC G06Q 30/0204 (2013.01) [G06N 3/042 (2023.01); G06N 5/02 (2013.01); G06Q 30/0255 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method utilizing machine learning models for data segmentation in an industry, comprising:
receiving, by a device, user data identifying purchases by users of user devices and identifying non-temporal data associated with the users;
preprocessing, by the device, the user data to generate sequences of multivariate and multimodal symbols;
processing, by the device, the sequences of multivariate and multimodal symbols, with a long short-term memory based encoder-decoder model, to generate sequence embeddings,
wherein the sequence embeddings are a transformation of the sequences into embeddings of fixed length,
wherein the embeddings of the fixed length are an internal representation of concepts within a neural network and include vectors of numbers that are learned by the neural network from input data during a training stage, and
wherein the sequence embeddings are generated based on initializing hidden vectors of the long short-term memory based encoder-decoder model and updating the hidden vectors with the sequences of multivariate and multimodal symbols;
converting, by the device, the non-temporal data associated with the users, to a knowledge graph, to determine knowledge graph embeddings capturing the non-temporal data associated with the users,
wherein the knowledge graph embeddings represent the non-temporal data in a continuous vector space to predict missing links between nodes of the knowledge graph;
training, by the device, the sequence embeddings and the knowledge graph embeddings jointly, to generate modified sequence embeddings that capture both temporal data and the non-temporal data;
processing, by the device, the sequence embeddings and the knowledge graph embeddings, with a knowledge graph embedding model;
generating, by the device and based on processing the sequence embeddings and the knowledge graph embeddings with the knowledge graph embedding model, the modified sequence embeddings, wherein the knowledge graph embedding model modifies the sequence embeddings by processing output data provided by the knowledge graph embeddings;
processing, by the device, the modified sequence embeddings, with a clustering model, to determine clusters of the users in relation to products or services purchased by the users; and
performing, by the device and based on the clusters of the users, one or more actions comprising at least one of:
retraining at least one of the long short-term memory based encoder-decoder model, the knowledge graph embedding model, or the clustering model based on the determined clusters of the users,
modifying at least one of the products or the services based on the determined clusters, or
generating and providing marketing data for the products or the services to user devices associated with the clusters of the users.