Decision Tree Splitting Criterion for Web Content Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine-learning systems for personalizing web content face challenges in execution speed and storage requirements due to the use of large prediction models with thousands of decision trees, making it computationally infeasible to determine customized content in real-time for users.
Innovation Solution
The implementation of prediction models using decision trees generated with a splitting criterion that requires a minimum number of positive and negative instances on each side of a split, reducing the depth and number of nodes in the decision trees, thereby reducing processing time and storage needs while maintaining prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If large prediction models with thousands of decision trees are used to improve prediction accuracy, then prediction accuracy is improved, but processing time increases and storage requirements increase
Solution Approach 1:
The patent changes the parameters of decision tree generation by introducing minimum instance requirements (minimum number of positive and negative instances) at each split node. This parameter modification prunes unnecessary branches and reduces the overall size of decision trees while maintaining prediction accuracy, thereby resolving the contradiction between accuracy and processing time
Solution Approach 2:
The patent applies partial action by not building complete decision trees with all possible splits, but rather stopping at splits that satisfy minimum instance requirements. This partial construction approach reduces computational complexity and storage needs while preserving the essential predictive capability
2Measurement precision
If large prediction models with thousands of decision trees are used to improve prediction accuracy, then prediction accuracy is improved, but storage space requirements increase
Solution Approach 1:
The patent modifies the parameters of decision tree construction by enforcing minimum instance requirements at each split, which significantly reduces the number of nodes and branches. This parameter change leads to more compact decision trees that consume less storage space while maintaining predictive accuracy
Solution Approach 2:
The patent extracts and removes unnecessary branches from decision trees that do not meet minimum instance requirements. By taking out these redundant portions, the overall model size is reduced, decreasing storage requirements while preserving the core predictive functionality
3Measurement precision
If prediction models with many decision trees are used to improve prediction accuracy, then prediction accuracy is improved, but computational feasibility decreases
Solution Approach 1:
The patent changes the computational parameters by introducing minimum instance thresholds that control the growth of decision trees. This parameter modification makes the computational process more feasible by limiting the number of operations required during model construction and execution, while still achieving accurate predictions
Data Source
AI summary
In one aspect, a request for web content is received from a user device communicatively coupled to the processing device via the network. In response to receiving the request, user information associated with the user is determined. Predicted responses of the user to each variation of a plurality of variations of the web content are determined using prediction models and the user information. The prediction models include one or more decision trees generated using a splitting criterion requiring a minimum number of positive responses to a variation and a minimum number of negative responses to the variation as a condition of considering the possible split. The variation determined to have a threshold likelihood of yielding a predicted positive response of the predicted responses is selected based on the user information. The variation is transmitted to the user device via the network.