Database-Based Decision Tree Training with Parallel Variance Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Training decision trees and random forests is computationally expensive due to the need for extensive calculations and resource-intensive processes, particularly when dealing with large datasets and numerous features.
Innovation Solution
Utilize a database system to store input data and employ variance-based techniques to select and associate decision nodes, allowing for parallel processing and reduced computational overhead by calculating variance for each feature to determine the highest variance feature for branching.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional decision tree training methods are used with extensive calculations on large datasets, then accurate classification results are achieved, but computational cost and training time increase significantly
Solution Approach 1:
The patent segments the feature set into multiple subsets and processes them in parallel using multiple threads. Each thread calculates variance for a specific feature subset independently, allowing concurrent processing that reduces overall training time while maintaining classification accuracy through comprehensive feature analysis.
Solution Approach 2:
The patent performs preliminary variance calculation on training data to identify the highest variance feature before building the decision tree. This preliminary action allows the system to pre-determine optimal branching features, reducing the computational burden during actual tree construction and classification operations.
2Measurement precision
If traditional decision tree training methods are used with extensive calculations on large datasets, then accurate classification results are achieved, but computational resources and processing power are consumed excessively
Solution Approach 1:
The patent divides the computationally intensive variance calculation task into multiple segments processed in parallel. By splitting the feature set across multiple threads, each thread handles a manageable subset with reduced memory access contention and computational overhead, achieving better resource utilization while maintaining overall accuracy.
Solution Approach 2:
The system uses the database system's native parallel query execution capabilities to perform variance calculations. The database engine leverages its own optimized execution plans and resource management to handle the computational tasks efficiently, reducing the burden on external processing systems.
3Measurement precision
If variance calculations are performed for all features sequentially, then the highest variance feature is identified accurately, but the training process becomes time-consuming
Solution Approach 1:
The patent segments the feature calculation tasks and assigns them to multiple threads that execute simultaneously. Each thread calculates variance for its assigned features independently, and the results are aggregated to identify the highest variance feature. This parallel approach maintains selection accuracy while dramatically improving training efficiency through concurrent execution.
Data Source
AI summary
In an embodiment, a computer-implemented method for training a decision tree using a database system, the decision tree comprising a plurality nodes, comprises, by one or more computing devices: storing in a database input data for training the decision tree, the input data comprising a plurality of feature values corresponding to a plurality of features; generating a particular node of the plurality of decision nodes by: selecting a subset of the plurality of features and a subset of the input data; using one or more queries to the database system, for each feature of the subset of the plurality of features, calculating a variance value associated with the feature based on the subset of the input data; identifying a particular feature of the subset of the plurality of features associated with the highest variance value; associating the particular node with the particular feature, wherein the particular node causes the decision tree to branch based on the particular feature.