Co-clustering Prediction Model for Customer Merchandise Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing prediction models for customer-merchandise relationships in marketing have low accuracy due to failure to account for differences in customer characteristics beyond attribute values, leading to inaccurate purchase probability predictions.
Innovation Solution
A prediction model generation system that performs co-clustering on customer and merchandise IDs based on master data and fact data, using co-clustering processing and prediction model generation to refine clusters and improve prediction accuracy by adjusting membership probabilities based on attribute value differences and error calculations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a prediction model is generated using only attribute values from master data, then the model can be constructed with available data, but the prediction accuracy is low due to failure to account for differences in customer characteristics beyond attribute values
Solution Approach 1:
The patent segments customers and merchandise into distinct clusters based on their relationships and attributes. Customer clustering groups customers with similar purchasing patterns, while merchandise clustering groups merchandise with similar characteristics. This segmentation allows the model to capture nuanced differences within segments that attribute-based models miss, thereby improving prediction accuracy without requiring fundamentally new data sources.
Solution Approach 2:
The patent introduces a new dimension of analysis by creating cluster assignments as an additional layer of information. Instead of relying solely on traditional attribute dimensions (age, gender, income), the model adds cluster membership dimensions that represent latent patterns in purchasing behavior. This dimensional expansion enables the model to distinguish between customers with identical attributes but different behavioral patterns, improving prediction accuracy.
2Measurement precision
If co-clustering processing is performed on customer and merchandise IDs, then the prediction accuracy is improved by reflecting actual characteristics, but the computational complexity and processing time increase
Solution Approach 1:
The patent performs preliminary clustering operations to group customers and merchandise before generating the final prediction model. By pre-organizing data into clusters based on historical purchasing patterns, the system reduces the computational burden during model training and inference. This preliminary structuring allows for faster processing when making predictions, as the model can leverage pre-computed cluster relationships rather than analyzing all customer-merchandise pairs from scratch.
Solution Approach 2:
The co-clustering algorithm automatically discovers and organizes patterns in the data without requiring manual intervention or extensive parameter tuning. The system self-organizes customers and merchandise into meaningful clusters based on their inherent relationships, reducing the need for complex preprocessing and feature engineering. This self-organizing capability improves processing efficiency by eliminating time-consuming manual data preparation steps.
3Measurement precision
If clusters are refined by adjusting membership probabilities based on error calculations, then the prediction model accuracy is enhanced, but the computational resources and processing steps increase
Solution Approach 1:
The patent implements an iterative refinement process where prediction errors feed back into cluster membership adjustments. The model generates initial predictions, calculates errors between predicted and actual purchasing behavior, and uses these errors to adjust cluster membership probabilities. This feedback loop continuously improves cluster quality and prediction accuracy. The systematic use of error signals to guide refinement reduces the need for trial-and-error approaches, making the complex processing more manageable and efficient.
Data Source
AI summary
A prediction model generation system is provided that is capable of generating a prediction model for accurately predicting a relationship between an ID of a record in first master data and an ID of a record in second master data. Co-clustering means 71 performs co-clustering processing for performing co-clustering on first IDs and second IDs in accordance with first master data, second master data, and fact data indicating a relationship between each of the first IDs and each of the second IDs. Prediction model generation means 72 performs prediction model generation processing for generating a prediction model for each combination of a first ID cluster and a second ID cluster. The prediction model uses the relationship between each of the first IDs and each of the second IDs as an objective variable. The first ID cluster serves as a cluster of the first IDs. The second ID cluster serves as a cluster of the second IDs. The prediction model generation processing and the co-clustering processing are repeated until it is determined that a prescribed condition is satisfied.