Abstract
Knowledge generalization of ruled-based models, such as decision trees or decision rules, have emerged from different backgrounds. This particular kind of models, given their interpretability, offer several possibilities to be combined. Despite each distinct context, common patterns have emerged revealing the systemic nature of the problem. In this paper, we look at the problem of generalizing the knowledge contained in a set of models as a process formalizing the operations that can be addressed in alternative ways. We also include a set-up to evaluate gen-eralized models based on their ability to replace the base ones from a predictive performance perspective, without loss of interpretability.

Abstract
The increasingly aging society in developed countries has raised attention to the role of technology in seniors' lives, namely concerning isolation-related issues. Independent seniors that live alone frequently neglect meals, hydration and proper medication-taking behavior. This work aims at eating and drinking recognition in free-living conditions for triggering smart reminders to autonomously living seniors, keeping system design considerations, namely usability and senior-acceptance criteria, in the loop. To that end, we conceived a new dataset featuring accelerometer and gyroscope wrist data to conduct the experiments. We assessed the performance of a single multi-class classification model when compared against several binary classification models, one for each activity of interest (eating vs. non-eating; drinking vs. non-drinking). Binary classification models performed consistently better for all tested classifiers (k-NN, Naive Bayes, Decision Tree, Multilayer Perceptron, Random Forests, HMM). This evidence supported the proposal of a semi-hierarchical activity recognition algorithm that enabled the implementation of two distinct data stream segmentation techniques, the customization of the classification models of each activity of interest and the establishment of a set of restrictions to apply on top of the classification output, based on daily evidence. An F1-score of 97% was finally attained for the simultaneous recognition of eating and drinking in an all-day acquisition from one young user, and 93% in a test set with 31 h of data from 5 different unseen users, 2 of which were seniors. These results were deemed very promising towards solving the problem of food and fluids intake monitoring with practical systems which shall maximize user-acceptance.

Abstract
Smartphones are increasingly present in human’s life. For example, for entertainment many people use their smartphones to watch videos or listen to music. Many users, however, stream or play videos with the intention to only listen to the audio track. This way, some battery energy, which is critical to most users, is unnecessarily consumed thus and switching between video and audio can increase the time of use of the smartphone between battery recharges. In this paper, we present a first approach that, based on the user context, can automatically switch between video and audio. A supervised learning approach is used along with the classifiers K-Nearest Neighbors, Hoeffding Trees and Naive Bayes, individually and combined to create an ensemble classifier. We investigate the accuracy for recognizing the context of the user and the overhead that this system can have on the smartphone energy consumption. We evaluate our approach with several usage scenarios and an average accuracy of 88.40% was obtained for the ensemble classifier. However, the actual overhead of the system on the smartphone energy consumption highlights the need for researching further optimizations and techniques. © 2019, Springer Nature Switzerland AG.

Abstract
The Classifier kNN is largely used in Human Activity Recognition systems. Research efforts have proposed methods to decrease the high computational costs of the original kNN by focusing, e.g., on approximate kNN solutions such as the ones relying on Locality-sensitive Hashing (LSH). However, embedded kNN implementations need to address the target device memory constraints and power/energy consumption savings. One of the important aspects is the constraint regarding the maximum number of instances stored in the kNN learning process (being it offline or online and incremental). This paper presents simple, energy/computationally efficient and real-time feasible schemes to maintain a maximum number of learning instances stored by kNN. Experiments in the context of HAR show the efficiency of our best approaches, and their capability to avoid the kNN storage runs out of training instances for a given activity, a situation not prevented by typical default schemes.

Abstract
Nowadays most people carry a smartphone with built-in sensors (e.g., accelerometers, gyroscopes) capable of providing useful data for Human Activity Recognition (HAR). Machine learning classification methods have been intensively researched and developed for HAR systems, each with different accuracy and performance levels. However, acquiring sensor data and executing machine learning classifiers require computational power and consume energy. As such, a number of factors, such as inadequate preprocessing, can have a negative impact on the overall HAR performance, even on high-end handheld devices. While high accuracy can be extremely important in some applications, the device’s battery life can be highly critical to the end-user. This paper is focused on the k-nearest neighbors’ algorithm (kNN), one of the most used algorithms in HAR systems, and research and develop energy-efficient implementations for mobile devices. We focus on a kNN implementation based on Locality-Sensitive Hashing (LSH) with a significant positive impact on the device’s battery life, fully integrated into a mobile HAR Android application able to classify human activities in real-time. The proposed kNN implementation was able to achieve execution time reductions of 50% over other versions of kNN with average accuracy of 96.55% when considering 8 human activities. © 2019, Springer Nature Switzerland AG.

Abstract
In data streams new classes can appear over time due to changes in the data statistical distribution. Consequently, models can become outdated, which requires the use of incremental learning algorithms capable of detecting and learning the changes over time. However, when a single classification model is used for novelty detection, there is a risk that its bias may not be suitable for new data distributions. A solution could be the combination of several models into an ensemble. Besides, because models can only be updated when labeled data arrives, we propose two unsupervised ensemble approaches: one combining clustering partitions using the same clustering technique; and other using different clustering techniques. We compare the performance of the proposed methods with well known novelty detection algorithms. The methods were tested on datasets commonly used in the novelty detection literature. The experimental results show that proposed ensembles have competitive performance for novelty detection in data streams. © Springer Nature Switzerland AG 2019.