The ARIEN Project is pleased to announce the publication of a new scientific contribution entitled \u201cLTC-CPD+: A Neural-based Framework for Online Change Point Detection in Multivariate Time Series.\u201d<\/strong><\/p>\n The study addresses the challenge of identifying change points<\/strong> in continuously evolving multivariate data. A change point occurs when the statistical characteristics of a time series\u2014such as its average values, variability, correlations or seasonal patterns\u2014shift significantly. Detecting such changes rapidly is particularly important in contexts where large quantities of data must be continuously analysed and timely responses are required.<\/p>\n The proposed LTC-CPD+ framework<\/strong> introduces a neural-based approach designed to detect changes while data are being generated. It uses Liquid Time-Constant networks as recurrent encoders and combines them with online selective inference, robust reconstruction losses and data-augmentation techniques specifically adapted to streaming environments. These elements are intended to improve the model\u2019s robustness, stability and capacity to adapt when only a limited number of observations are available after a change has occurred.<\/p>\n The framework was evaluated on four real-world benchmark datasets. The results show competitive or superior performance compared with the baseline methods considered in the study, with particularly strong improvements in datasets containing multiple change points.<\/p>\n This research contributes to ARIEN\u2019s broader work on advanced Artificial Intelligence and data-analysis methodologies. The capacity to identify emerging changes and unusual patterns in complex data streams can support the development of more responsive and reliable analytical systems, strengthening the technological foundations needed to address evolving security and organised-crime challenges.<\/p>\n The contribution is currently available as an open-access preprint and is expected to be published in the proceedings of the International Conference on Web Intelligence, Machine Intelligence and Semantics \u2013 WIMS 2026<\/strong>, taking place in Thessaloniki, Greece, from 11 to 14 October 2026. The available preprint has not yet undergone peer review or subsequent editorial corrections.<\/p>\n