A CMAC-based scheme for determining membership with classification of text strings

Area of Science:

• Computer Science
• Data Mining
• Machine Learning

Background:

• Membership determination of text strings is crucial for analyzing large textual datasets, especially under time constraints.
• Bloom filters offer a succinct structure for membership determination but parallel implementations increase false-positive errors.
• Classifying text strings alongside membership determination is increasingly desirable but challenging with existing methods.

Purpose of the Study:

• To propose a novel scheme for efficient text string membership determination and classification.
• To reduce false-positive errors inherent in parallel Bloom filter approaches.
• To enable simultaneous membership and classification using a single-layer calculation.

Main Methods:

• A new scheme based on Cerebellar Model Articulation Controller (CMAC), a neural network mapping, is proposed.
• A specialized hash function for text strings is developed to work with the CMAC scheme.
• Membership and classification are performed through a single-layer neural network calculation.

Main Results:

• The proposed CMAC-based scheme significantly reduces false-positive errors compared to parallel Bloom filters.
• The scheme effectively converges membership acceptance ranges for each class, enhancing classification accuracy.
• Simulations confirm superior performance with identical memory usage across different classification levels.

Conclusions:

• The proposed CMAC-based scheme offers a more accurate and efficient solution for text string membership determination with classification.
• This approach effectively mitigates the false-positive error issue prevalent in parallel Bloom filters.
• The method provides a robust alternative for large-scale textual data analysis requiring both membership and classification.