Abstract
Predictive maintenance methods assist early detection of failures and errors in machinery before they reach critical stages. This study proposes a data-driven predictive maintenance framework for the air production unit (APU) system of a train of Metro do Porto by deep learning based on a sparse autoencoder (SAE) network that efficiently detects abnormal data and considerably reduces the false alarm rate. Several analog and digital sensors installed on the APU system allow the detection of behavioral changes and deviations from the normal pattern by analyzing the collected data. We implemented two versions of the SAE network in which we inputted analog sensors data and digital sensors data, and the experimental results show that the failures due to air leakage problems are predicted by analog sensors data while other types of failures are identified by digital sensors data. A low pass filter is applied to the output of the SAE network, and a sequence of abnormal data is used as an alarm for the APU system failure. Performance indicators of the SAE network with digital sensors data, in terms of F1 Score, Recall, and Precision, are respectively, about 33.6%, 42%, and 28% better than those of the SAE network with analog sensors data. For comparison purposes, we also implemented a variational autoencoder (VAE). The results show that SAE performance is better than that of VAE by 14%, 77%, and 37% respectively, for Recall, Precision and F1 Score.

Abstract
There has been a significant effort by the research community to address the problem of providing methods to organize documentation, with the help of Information Retrieval methods. In this paper, we present several experiments with stream analysis methods to explore streams of text documents. This paper also presents possible architectures of the Text Document Stream Organization, with the use of incremental algorithms like Incremental Sparse TF-IDF and Incremental Similarity. Our results show that with this architecture, significant improvements are achieved, regarding efficiency in grouping of similar documents. These improvements are important since it is of general knowledge that great amounts of text analysis are a high dimensional and complex subject of study, in the data analysis area.

Abstract
This article presents our recent work on the topic of learning from data streams. We focus on emerging topics, including fraud detection, learning from rare cases, and hyper-parameter tuning for streaming data. © 2021, Springer Nature Switzerland AG.

Abstract
We present an online optimization method for time-evolving data streams that can automatically adapt the hyper-parameters of an embedding model. More specifically, we employ the Nelder-Mead algorithm, which uses a set of heuristics to produce and exploit several potentially good configurations, from which the best one is selected and deployed. This step is repeated whenever the distribution of the data is changing. We evaluate our approach on streams of real-world as well as synthetic data, where the latter is generated in such way that its characteristics change over time (concept drift). Overall, we achieve good performance in terms of accuracy compared to state-of-the-art AutoML techniques.

Abstract
In recent years data stream mining and learning from imbalanced data have been active research areas. Even though solutions exist to tackle these two problems, most of them are not designed to handle challenges inherited from both problems. As far as we are aware, the few approaches in the area of learning from imbalanced data streams fall in the context of classification, and no efforts on the regression domain have been reported yet. This paper proposes a technique that uses sampling strategies to cope with imbalanced data streams in a regression setting, where the most important cases have rare and extreme target values. Specifically, we employ under-sampling and over-sampling strategies that resort to Chebyshev's inequality value as a heuristic to disclose the type of incoming cases (i.e. frequent or rare). We have evaluated our proposal by applying it in the training of models by four well-known regression algorithms over fourteen benchmark data sets. We conducted a series of experiments with different setups on both synthetic and real-world data sets. The experimental results confirm our approach's effectiveness by showing the models' superior performance trained by each of the sampling strategies compared with their baseline pairs.

Abstract
Topic modeling or inference has been one of the well-known problems in the area of text mining. It deals with the automatic categorisation of words or documents into similarity groups also known as topics. In most of the social media platforms such as Twitter, Instagram, and Facebook, hashtags are used to define the content of posts. Therefore, modelling of hashtags helps in categorising posts as well as analysing user preferences. In this work, we tried to address this problem involving hashtags that stream in real-time. Our approach encompasses graph of hashtags, dynamic sampling and modularity based community detection over the data from a popular social media engagement application. Further, we analysed the topic clusters' structure and quality using empirical experiments. The results unveil latent semantic relations between hashtags and also show frequent hashtags in a cluster. Moreover, in this approach, the words in different languages are treated synonymously. Besides, we also observed top trending topics and correlated clusters.