Multi-Encoder Model for Customer Attrition Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Processing large amounts of data to predict customer attrition in financial institutions requires significant computer resources, and existing pattern recognition systems struggle to accurately account for behavior over time and static attributes without requiring extensive feature engineering and iterative pattern comparisons.
Innovation Solution
A machine learning model with multiple encoders and prediction layers that generates feature vectors for transactions and account attributes, producing embeddings that are concatenated and propagated through prediction layers to predict customer attrition, allowing for more accurate and efficient processing of time-series transactional and static data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If statistical algorithms compare transaction history of each customer to different patterns, then attrition prediction accuracy is improved, but computational resources required increase significantly
Solution Approach 1:
The patent segments the customer data processing into distinct feature extraction and pattern matching components. Feature vectors are extracted from transaction histories and customer attributes separately, then compared against stored patterns. This segmentation allows for more efficient processing by pre-computing and storing characteristic patterns, reducing the computational burden of real-time analysis while maintaining prediction accuracy.
2Adaptability or versatility
If more patterns are stored to account for customers with different characteristics, then prediction coverage is improved, but device complexity increases
Solution Approach 1:
The patent creates simplified copies of customer behavior patterns by extracting essential features and storing them as feature vectors. Instead of storing complete transaction histories or complex behavioral models, the system stores condensed feature representations that capture the essential characteristics of different customer segments. This copying approach enables the system to handle diverse customer characteristics with reduced complexity.
3Measurement precision
If iterative pattern comparisons are performed for each customer, then prediction accuracy is improved, but processing time increases
Solution Approach 1:
The patent performs preliminary actions by pre-extracting feature vectors from customer transaction histories and attributes before the actual prediction process. Characteristic patterns are identified and stored in advance, allowing the system to quickly compare new customer data against these pre-computed patterns. This preliminary processing significantly reduces the time required for iterative comparisons while maintaining prediction accuracy.
Data Source
AI summary
A method comprises generating a first feature vector comprising a plurality of values for a plurality of transactions, each of the plurality of transactions corresponding to an account and performed within a defined time period, and a second feature vector comprising an attribute value of an attribute of the account; inserting the first feature vector into a first encoder of a machine learning model to generate a transaction embedding and the second feature vector into a second encoder of the machine learning model to generate an attribute embedding; concatenating the transaction embedding and the attribute embedding to generate a concatenated embedding; and generating an account prediction value by propagating the concatenated embedding into a set of prediction layers of the machine learning model.