Rafael Rocha

Antonio J. G. Busson, Rafael Rocha, Rennan Gaio et al.

This work proposes the Two-headed DragoNet, a Transformer-based model for hierarchical multi-label classification of financial transactions. Our model is based on a stack of Transformers encoder layers that generate contextual embeddings from two short textual descriptors (merchant name and business activity), followed by a Context Fusion layer and two output heads that classify transactions according to a hierarchical two-level taxonomy (macro and micro categories). Finally, our proposed Taxonomy-aware Attention Layer corrects predictions that break categorical hierarchy rules defined in the given taxonomy. Our proposal outperforms classical machine learning methods in experiments of macro-category classification by achieving an F1-score of 93\% on a card dataset and 95% on a current account dataset.

Ruy Luiz Milidiú, Rafael Henrique Santos Rocha

The goal of Ordinal Regression is to find a rule that ranks items from a given set. Several learning algorithms to solve this prediction problem build an ensemble of binary classifiers. Ranking by Projecting uses interdependent binary perceptrons. These perceptrons share the same direction vector, but use different bias values. Similar approaches use independent direction vectors and biases. To combine the binary predictions, most of them adopt a simple counting heuristics. Here, we introduce a novel cumulative sum scoring function to combine the binary predictions. The proposed score value aggregates the strength of each one of the relevant binary classifications on how large is the item's rank. We show that our modeling casts ordinal regression as a Structured Perceptron problem. As a consequence, we simplify its formulation and description, which results in two simple online learning algorithms. The second algorithm is a Passive-Aggressive version of the first algorithm. We show that under some rank separability condition both algorithms converge. Furthermore, we provide mistake bounds for each one of the two online algorithms. For the Passive-Aggressive version, we assume the knowledge of a separation margin, what significantly improves the corresponding mistake bound. Additionally, we show that Ranking by Projecting is a special case of our prediction algorithm. From a neural network architecture point of view, our empirical findings suggest a layer of cusum units for ordinal regression, instead of the usual softmax layer of multiclass problems.