Abstract

Abstract
The problem of joint estimation of parameters and state of continuous time systems using discrete time observations is addressed. The plant parameters are assumed to be modeled by a Wiener process. The a priori probability density function (pdf) of an extended state that comprises the plant state variables and the parameters is propagated in time using an approximate solution of the Fokker-Planck equation that relies on Trotter's formula for semigroup decomposition. The a posteriori (i. e., given the observations) pdf is then computed at the observation instants using Bayes law.

Abstract
In this work we study a way to explore and extract more information from data sets with a hierarchical tree structure. We propose that any statistical study on this type of data should be made by group, after clustering. In this sense, the most adequate approach is to use the Mahalanobis-Wasserstein distance as a measure of similarity between the cases, to carry out clustering or unsupervised classification. This methodology allows for the clustering of cases, as well as the identification of their profiles, based on the distribution of all the variables that characterises each subject associated with each case. An application to a set of teenagers' interviews regarding their habits of communication is described. The interviewees answered several questions about the kind of contacts they had on their phone, Facebook, email or messenger as well as the frequency of communication between them. The results indicate that the methodology is adequate to cluster this kind of data sets, since it allows us to identify and characterise different profiles from the data. We compare the results obtained with this methodology with the ones obtained using the entire database, and we conclude that they may lead to different findings.

Abstract
Online incremental models for recommendation are nowadays pervasive in both the industry and the academia. However, there is not yet a standard evaluation methodology for the algorithms that maintain such models. Moreover, online evaluation methodologies available in the literature generally fall short on the statistical validation of results, since this validation is not trivially applicable to stream-based algorithms. We propose a k-fold validation framework for the pairwise comparison of recommendation algorithms that learn from user feedback streams, using prequential evaluation. Our proposal enables continuous statistical testing on adaptive-size sliding windows over the outcome of the prequential process, allowing practitioners and researchers to make decisions in real time based on solid statistical evidence. We present a set of experiments to gain insights on the sensitivity and robustness of two statistical tests-McNemar's and Wilcoxon signed rank-in a streaming data environment. Our results show that besides allowing a real-time, fine-grained online assessment, the online versions of the statistical tests are at least as robust as the batch versions, and definitely more robust than a simple prequential single-fold approach.

Abstract

Abstract
As AI algorithms thrive on data, SCADA would be considered a natural ground for Artificial Intelligence (AI) applications to be developed, translating that avalanche of information into meaningful and fast insights to human operators. However, presently, the high complexity of the events, the data semantics, the large variety of equipment and technologies translate into very few AI applications developed in SCADA. Aware of the enormous potential yet to be explored, E-REDES partnered with INESC TEC to experiment on the development of two novel AI applications based on SCADA data. The first tool, called Alarm2Insights, identifies anomalous behaviours regarding the performance of the protection functions associated with HV and MV line panels. The second tool, called EventProfiler, uses unsupervised learning to identify similar events (i.e., with similar log messages) in HV line panels, and supervised learning to classify new events into previously defined clusters and detect unique or rare events. Aspects associated to data handling and pre-processing are also discussed. The project's results show a very promising potential of applying AI to SCADA data, enhancing the role of the operator and support him in doing better and more informed decisions. © 2021 The Institution of Engineering and Technology.