Attention Augmented Convolutional Transformer for Tabular Time-series

TL;DR

This paper introduces a scalable, end-to-end transformer-based architecture for tabular time-series classification, incorporating novel masking, timestamp embedding, and convolutional techniques to effectively model periodic and aperiodic patterns.

Contribution

It presents a novel architecture combining transformers and convolutions with new timestamp embedding for improved tabular time-series representation learning.

Findings

Effective handling of periodic and aperiodic patterns.

No need for quantization of continuous features.

Improved performance on classification tasks.

Abstract

Time-series classification is one of the most frequently performed tasks in industrial data science, and one of the most widely used data representation in the industrial setting is tabular representation. In this work, we propose a novel scalable architecture for learning representations from tabular time-series data and subsequently performing downstream tasks such as time-series classification. The representation learning framework is end-to-end, akin to bidirectional encoder representations from transformers (BERT) in language modeling, however, we introduce novel masking technique suitable for pretraining of time-series data. Additionally, we also use one-dimensional convolutions augmented with transformers and explore their effectiveness, since the time-series datasets lend themselves naturally for one-dimensional convolutions. We also propose a novel timestamp embedding technique, which helps in handling both periodic cycles at different time granularity levels, and aperiodic trends present in the time-series data. Our proposed model is end-to-end and can handle both categorical and continuous valued inputs, and does not require any quantization or encoding of continuous features.