Multi-class Decision Tree Model for Audience Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional systems for generating digital audience segments are inaccurate, inefficient, and inflexible, often resulting in significant overlap between segments, leading to misclassification of client devices and wasteful distribution of digital content.
Innovation Solution
The implementation of a multi-class decision tree machine-learning model that utilizes a customized penalty loss matrix, generated through regression models based on reach and accuracy metrics, to classify client devices into non-overlapping audience segments, with the ability to adjust tree depth to improve accuracy and avoid overfitting.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional systems are used to generate digital audience segments, then the process is simple, but the accuracy is poor and segments have large overlap
Solution Approach 1:
The patent applies segmentation by dividing the audience classification problem into multiple distinct audience segments using a multi-class decision tree model. Each leaf node in the decision tree represents a separate audience segment with specific classification rules, enabling precise segmentation of client devices into non-overlapping groups based on their traits and behaviors.
Solution Approach 2:
The patent implements dynamics through auto-tunable parameters including dynamic adjustment of tree depth, penalty loss matrix values, and reach/accuracy metrics. The system can adaptively modify these parameters based on performance feedback and requirements, allowing the classification model to optimize its structure and behavior dynamically rather than using fixed configurations.
2Productivity
If conventional systems generate audience segments, then computational resources are used, but the efficiency is low and overlap is large
Solution Approach 1:
The patent extracts and eliminates the harmful overlap between audience segments by using a decision tree structure that assigns each client device to exactly one leaf node segment. The exclusive classification mechanism removes the redundancy and waste associated with conventional systems where devices could belong to multiple overlapping segments, thereby improving efficiency and reducing computational resource waste.
Solution Approach 2:
The system implements self-service through automated model training and parameter optimization. The multi-class decision tree model automatically learns classification rules from training data, performs self-adjustment of parameters, and generates audience segments without requiring manual intervention, thereby significantly improving productivity and reducing the need for extensive computational resources compared to manual segmentation approaches.
3Measurement precision
If tree depth is increased to improve accuracy, then classification accuracy improves, but overfitting occurs
Solution Approach 1:
The patent applies parameter changes by implementing auto-tunable tree depth and penalty loss matrix parameters that can be dynamically adjusted based on performance metrics. The system monitors classification accuracy and generalization performance, automatically modifying the tree depth parameter to prevent overfitting while maintaining high accuracy. This allows the model to adapt its complexity parameter to match the underlying data structure without manual intervention.
Solution Approach 2:
The system implements feedback mechanisms by evaluating model performance on validation data and using this feedback to adjust tree depth and other parameters. The auto-tuning process continuously monitors for signs of overfitting and adjusts the model complexity accordingly, ensuring that the classification accuracy improves while maintaining reliable generalization to unseen data.
4Measurement precision
If customized penalty loss matrix is used, then reach and accuracy metrics are optimized, but model training complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-computing and storing penalty loss matrices for different audience segment misclassifications before model training. These pre-computed matrices encode the reach and accuracy requirements for each segment, allowing the model to optimize for these metrics during training without requiring complex real-time calculations. This preliminary preparation simplifies the training process while maintaining optimization for reach and accuracy.
Data Source
AI summary
This disclosure describes one or more implementations of a model segmentation system that generates accurate audience segments for client devices/individuals utilizing multi-class decision tree machine-learning models. For example, in various implementations, the model segmentation system generates a customized loss penalty matrix from multiple loss penalty matrices. In particular, the model segmentation system can generate regression mappings of model evaluation metrics for a plurality of decision tree models and combine loss penalty matrices based on the regression mappings to generate a customized loss penalty matrix that best fits an administrator's customized needs of segment accuracy and reach. The model segmentation system then utilizes the customized loss penalty matrix to train a multi-class decision tree machine-learning model to classify client devices into non-overlapping audience segments. Further, in one or more implementations, the model segmentation system refines the multi-class decision tree machine-learning model based on adjusting the tree depth.