CoTraM: Convolutional Transformer for Multichannel Time-Series Classification

The computational analysis of multichannel time series has established its significance in a myriad of domains, spanning satellite data interpretation, environmental monitoring, and financial forecasting, to name a few. With the complexity and significant length of time series data, there arises an exigent need for advanced processing mechanisms. This is where the Convolutional-Transformer Model (CoTraM) makes its mark. Designed primarily for generalized multichannel time series classification, this architecture has a special aptitude for handling extremely lengthy sequences. The research at hand delves deep into CoTraM's adaptability and efficacy across diverse datasets. Of particular note is its efficiency in processing extended clinical sequences, such as Electroencephalograms (EEG) and Polysomnography data. The potential for CoTraM to serve as an instrumental aid to clinicians, who are often faced with the arduous task of analyzing lengthy data for prognostic insights, stands at the forefront of this investigation.