Feature Construction for Relational Sequence Learning

TL;DR

This paper introduces a feature construction and selection method for relational sequence learning, improving classification accuracy over traditional models like HMMs and CRFs by using a wrapper approach with a naive Bayes classifier.

Contribution

It presents a novel feature construction and selection framework specifically designed for multi-class relational sequence learning tasks.

Findings

Improved classification accuracy on real-world datasets.

Outperforms established methods like HMMs, Fisher kernels, and CRFs.

Demonstrates effectiveness of feature-based approach for relational sequences.

Abstract

We tackle the problem of multi-class relational sequence learning using relevant patterns discovered from a set of labelled sequences. To deal with this problem, firstly each relational sequence is mapped into a feature vector using the result of a feature construction method. Since, the efficacy of sequence learning algorithms strongly depends on the features used to represent the sequences, the second step is to find an optimal subset of the constructed features leading to high classification accuracy. This feature selection task has been solved adopting a wrapper approach that uses a stochastic local search algorithm embedding a naive Bayes classifier. The performance of the proposed method applied to a real-world dataset shows an improvement when compared to other established methods, such as hidden Markov models, Fisher kernels and conditional random fields for relational sequences.