Decision Tree Generation Using Weighted Graph Rule Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The decision-tree-based traffic classification method experiences a high occurrence probability of rule replication, leading to increased storage occupancy and inefficiencies in traffic classification processes.
Innovation Solution
A method involving encoding rules into encoded rule sets and generating weighted undirected graphs to reduce rule replication by combining vertices based on maximum or minimum weight thresholds, resulting in fewer rule subsets with lower replication probabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If rule sets are divided into multiple rule subsets using traditional cutting methods, then decision tree search efficiency is improved, but rule replication occurs frequently leading to increased storage occupancy
Solution Approach 1:
The patent applies parameter changes by transforming the rule representation from binary format to weighted undirected graph format. Each rule is encoded with weights representing the number of matching packets, and the cutting process uses these weight parameters to make more informed decisions about where to split rule sets, thereby reducing unnecessary rule replication while maintaining search efficiency.
Solution Approach 2:
The patent implements feedback mechanisms by calculating and using weight values (number of matching packets) to guide the cutting process. The system continuously evaluates the impact of potential cuts on rule replication and storage occupancy, adjusting the cutting strategy based on this feedback to minimize rule replication while preserving search performance.
2Quantity of substance
If more rule subsets are created to reduce rule replication, then storage occupancy decreases, but decision tree complexity and generation time increase
Solution Approach 1:
The patent transforms the rule set representation to include weight parameters that enable more efficient cutting strategies. By using weighted undirected graphs where edge weights represent rule relationships and packet matching counts, the system can identify optimal cut points that reduce storage occupancy without creating excessive decision tree complexity, as the weight information guides merges of compatible rules.
Solution Approach 2:
The patent applies merging by combining rules into weighted undirected graph structures where compatible rules can be grouped together. The cutting process merges rules that share common characteristics identified through weight analysis, reducing the overall number of rule subsets needed while maintaining low storage occupancy and avoiding excessive decision tree complexity.
3Ease of manufacture
If traditional cutting methods are used to divide rule sets, then decision tree generation is straightforward, but rule replication probability increases reducing classification efficiency
Solution Approach 1:
The patent changes the parameters used in decision tree generation from simple binary rule matching to weighted graph-based matching. By incorporating weight parameters that represent packet matching counts and rule relationships, the system maintains relatively straightforward generation processes while significantly reducing rule replication probability, thereby improving classification efficiency without sacrificing ease of generation.
Solution Approach 2:
The patent introduces weighted undirected graphs as an intermediary structure between the rule set and the decision tree. This intermediary representation with weight parameters enables more intelligent cutting decisions that reduce rule replication, while the graph structure itself serves as a bridge that maintains the relative simplicity of decision tree generation processes.
Data Source
AI summary
Embodiments of the present invention provide a method and an apparatus for generating a decision tree. The method includes generating an encoded rule set according to a rule set, generating a first weighted undirected graph, and calculating a weight of each side in the first weighted undirected graph. If a weight of a side with a maximum weight in the first weighted undirected graph is greater than a first threshold, a first operation is cyclically performed until a weight of a side with a maximum weight in a newly generated weighted undirected graph is less than or equal to the first threshold and a decision tree is generated for a rule subset corresponding to each vertex respectively in the newly generated weighted undirected graph.