Abstract
Performance estimation aims at estimating the loss that a predictive model will incur on unseen data. This process is a fundamental stage in any machine learning project. In this paper we study the application of these methods to time series forecasting tasks. For independent and identically distributed data the most common approach is cross-validation. However, the dependency among observations in time series raises some caveats about the most appropriate way to estimate performance in this type of data. Currently, there is no consensual approach. We contribute to the literature by presenting an extensive empirical study which compares different performance estimation methods for time series forecasting tasks. These methods include variants of cross-validation, out-of-sample (holdout), and prequential approaches. Two case studies are analysed: One with 174 real-world time series and another with three synthetic time series. Results show noticeable differences in the performance estimation methods in the two scenarios. In particular, empirical experiments suggest that blocked cross-validation can be applied to stationary time series. However, when the time series are non-stationary, the most accurate estimates are produced by out-of-sample methods, particularly the holdout approach repeated in multiple testing periods.

Abstract
Concept drift detection is a crucial task in data stream evolving environments. Most of the state of the art approaches designed to tackle this problem monitor the loss of predictive models. Accordingly, an alarm is launched when the loss increases significantly, which triggers some adaptation mechanism (e.g. retrain the model). However, this modus operandi falls short in many real-world scenarios, where the true labels are not readily available to compute the loss. These often take up to several weeks to be available. In this context, there is increasing attention to approaches that perform concept drift detection in an unsupervised manner, i.e., without access to the true labels. We propose a novel approach to unsupervised concept drift detection, which is based on a student-teacher learning paradigm. Essentially, we create an auxiliary model (student) to mimic the behaviour of the main model (teacher). At run-time, our approach is to use the teacher for predicting new instances and monitoring the mimicking loss of the student for concept drift detection. In a set of controlled experiments, we discovered that the proposed approach detects concept drift effectively. Relative to the gold standard, in which the labels are immediately available after prediction, our approach is more conservative: it signals less false alarms, but it requires more time to detect changes. We also show the competitiveness of our approach relative to other unsupervised methods. © 2020, Springer Nature Switzerland AG.

Abstract
Patients with end-stage renal disease should be educated and trained to take care of their own arteriovenous fistula (AVF) with the purpose of developing self-care behaviors concerning vascular access. This was a prospective and observational study. We designed this research to identify clinically meaningful self-care behavior profiles in hemodialysis (HD) patients, and it was carried out in a private dialysis unit in the Lisbon region, Portugal, involving 101 patients. The proportion of male patients was 66.3%, the mean age was 60.9 years, and the frequency of self-care behaviors was 71%. Cluster analysis based on the subscale scores grouped patients in two clusters named "moderate self-care" and "high self-care." Those profiles exhibit significant differences concerning gender, education, employment, dialysis vintage, AVF duration, and information on care with the AVF. Identification of self-care-behavior profiles in HD patients with AVF enables one to adjust education programs to the patients' characteristics.

Abstract

Abstract
Nuisance or unsolicited calls and instant messages come at any time in a variety of different ways. These calls would not only exasperate recipients with the unwanted ringing, impacting their productivity, but also lead to a direct financial loss to users and service providers. Telecommunication Service Providers (TSPs) often employ standalone detection systems to classify call originators as spammers or non-spammers using their behavioral patterns. These approaches perform well when spammers target a large number of recipients of one service provider. However, professional spammers try to evade the standalone systems by intelligently reducing the number of spam calls sent to one service provider, and instead distribute calls to the recipients of many service providers. Naturally, collaboration among service providers could provide an effective defense, but it brings the challenge of privacy protection and system resources required for the collaboration process. In this paper, we propose a novel decentralized collaborative system named privy for the effective blocking of spammers who target multiple TSPs. More specifically, we develop a system that aggregates the feedback scores reported by the collaborating TSPs without employing any trusted third party system, while preserving the privacy of users and collaborators. We evaluate the system performance of privy using both the synthetic and real call detail records. We find that privy can correctly block spammers in a quicker time, as compared to standalone systems. Further, we also analyze the security and privacy properties of the privy system under different adversarial models.

Abstract