Effective strategies for the prediction of aircraft passenger seat (dis-)comfort based on Time Series Classification

Aircraft seat is one of the factors which mostly impact on passenger’s flight experience and her/his willingness to choose the same airline in future occasions. The focus of this paper is on the prediction of passenger seat (dis-) comfort via objective methods based on the analysis of pressure distribution at seat interface. The aim is identifying the best strategy for predicting seat (dis-)comfort via the combination of well-known pressure indexes and Time Series Classification (TSC) algorithms applied in a univariate as well as a multivariate setting. To leverage the full potential of TSC algorithms, a comparison across different Data Augmentation (DA) techniques has been conducted. The comparison of seat (dis-)comfort prediction strategies provides useful insights on the informativeness of pressure features to accurately predict seat (dis-)comfort. Adopting a multivariate setting and expanding dataset size artificially have not enhanced predictive performance of TSC algorithms. The only exception results ResNet algorithm which in multivariate TSC benefits from DA, showing satisfactory predictive performance.