Search

Human Activity Recognition is a machine learning task for the classification of human physical activities. Applications for that task have been extensively researched in recent literature, specially due to the benefits of improving quality of life. Since wearable technologies and smartphones have become more ubiquitous, a large amount of information about a person’s life has become available. However, since each person has a unique way of performing physical activities, a Human Activity Recognition system needs to be adapted to the characteristics of a person in order to maintain or improve accuracy. Additionally, when smartphones devices are used to collect data, it is necessary to manage its limited resources, so the system can efficiently work for long periods of time. In this paper, we present a semi-supervised ensemble algorithm and an extensive study of the influence of hyperparameter configuration in classification accuracy. We also investigate how the classification accuracy is affected by the person and the activities performed. Experimental results show that it is possible to maintain classification accuracy by adjusting hyperparameters, like window size and window overlap, depending on the person and activity performed. These results motivate the development of a system able to automatically adapt hyperparameter settings for the activity performed by each person.

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 Consequently, models can become outdated, which requires the use of incremental learning algorithms capable of detecting and learning the changes over time.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.

Data streams are related to large amounts of data that can continuously arrive with a probability distribution that may change over time. Depending on the changes in the data distribution, different phenomena can occur, like new classes can appear or concept drift can occur in existing classes. Machine Learning algorithms have been often used to model this data. New classes are patterns that were not seen during the training of the current classification model, but appear after some time. Concept drift occurs when the concepts associated with a dataset change as new data arrive. This paper proposes a new algorithm based on kNN that uses micro-clusters as prototypes and incrementally updates the micro-clusters or creates new micro-clusters when novelties are detected. In the online phase, each instance close to a micro-cluster is considered an extension of the micro-cluster, being used to adapt the model to concept drift. The proposed algorithm is experimentally compared with a state-of-the-art classifier from the data stream literature and one baseline. According to the experimental results, the proposed algorithm increases the predictive performance over time by incrementally learning changes in the data distribution.

Data stream is a challenging research topic in which data can continuously arrive with a probability distribution that may change over time. Depending on the changes in the data distribution, different phenomena can occur, for example, a concept drift. A concept drift occurs when the concepts associated with a dataset change when new data arrive. This paper proposes a new method based on k-Nearest Neighbors that implements a sliding window requiring less instances stored for training than existing methods. For such, a clustering approach is used to summarize data by placing labeled instances considered similar in the same cluster. Besides, instances close to the uncertainty border of existing classes are also stored, in a sliding window, to adapt the model to concept drift. The proposed method is experimentally compared with state-of-the-art classifiers from the data stream literature, regarding accuracy and processing time. According to the experimental results, the proposed method has better accuracy and less time consumption when fewer information about the concepts are stored in a single sliding window.