Customer Segmentation via LSTM and Knowledge Graph Embeddings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current customer segmentation techniques are inefficient in processing non-temporal features and learning temporal features, leading to incorrect customer segments and ineffective marketing campaigns.
Innovation Solution
A method utilizing a long short-term memory based encoder-decoder model and a knowledge graph embedding model to generate sequence embeddings and knowledge graph embeddings, which are then processed with a clustering model to determine accurate user clusters based on purchase data and non-temporal data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If current customer segmentation techniques are used, then implementation is simple, but processing efficiency of non-temporal features and learning of temporal features is poor
Solution Approach 1:
The patent segments customer segmentation into two distinct processing streams: one for temporal features using LSTM-based encoder-decoder models and another for non-temporal features using knowledge graph embedding models. This segmentation allows each stream to be optimized for its specific data type, improving overall processing efficiency while maintaining manageable complexity through modular architecture.
Solution Approach 2:
The patent transforms non-temporal features into a new dimension by constructing knowledge graphs that represent relationships between customers, products, and attributes. This dimensional transformation allows non-temporal features to be processed differently from traditional tabular data, enabling more efficient processing and better capture of relational patterns.
2Measurement precision
If current customer segmentation techniques are used, then computing resource consumption is high, but segmentation accuracy is low
Solution Approach 1:
The patent replaces traditional mechanical data processing approaches with neural network-based models. Specifically, LSTM networks replace conventional time-series analysis methods, and knowledge graph embedding models replace traditional relational database queries. This substitution enables more accurate segmentation while optimizing computing resource usage through parallel processing capabilities of neural networks.
3Reliability
If current customer segmentation techniques are used, then implementation is straightforward, but marketing campaign effectiveness is low
Solution Approach 1:
The patent changes key parameters in the customer segmentation process by introducing sequence embeddings that capture temporal patterns in customer behavior and knowledge graph embeddings that capture relational structures. These parameter changes enable more accurate customer profiling, leading to improved marketing campaign effectiveness despite increased system complexity.
Solution Approach 2:
The patent introduces intermediate representations (embeddings) that serve as mediators between raw customer data and marketing decisions. Sequence embeddings act as intermediaries for temporal behavior patterns, while knowledge graph embeddings serve as intermediaries for relational features. These intermediaries enable more nuanced customer understanding and better-informed marketing strategies.
Data Source
AI summary
A device may receive purchase data identifying purchases by users of user devices and identifying non-temporal data associated with the users, and may preprocess the purchase data to generate sequences of multivariate and multimodal symbols. The device may process the sequences of multivariate and multimodal symbols, with a long short-term memory based encoder-decoder model, to generate sequence embeddings, and may process the non-temporal data associated with the users, with a knowledge graph, to determine knowledge graph embeddings capturing the non-temporal data. The device may process the sequence embeddings and the knowledge graph embeddings, with a knowledge graph embedding model, to generate modified sequence embeddings, and may process the modified sequence embeddings, with a clustering model, to determine clusters of the users in relation to products or services purchased by the users. The device may perform one or more actions based on the clusters of the users.