Fuels are crucial for any country's development and economy, impacting various sectors such as transportation, industry, and electricity generation. Accurate prediction of monthly fuel demand can improve supply chain management, strategic decision-making, and financial planning for businesses while helping governments develop decarbonization policies and estimate pollutant emissions. This paper explores machine learning models to forecast fossil fuels and biofuel demand 12 months ahead, using univariate time series data representing the historical sales of 27 Brazilian states, one of the world's leading producers and consumers of fuels. We evaluate different time series feature sets, machine learning regression models, and prediction strategies to address the complexity of fuel sales influenced by factors such as economic conditions and geopolitical events. Our comprehensive evaluation aims to determine an effective setting for predictive models in the fuel domain. Our results show that popular feature extractors for time series, such as Catch22 and TsFresh, cannot improve the original data representation for most forecasting models. Although focused on Brazil, our findings apply to other countries, since the trained models do not rely on external variables, such as micro and macroeconomic indicators.

Peaks in time series represent significant events in the measurements of a phenomenon over time, such as a sudden increase in the sales of an item on a specific day of the week. Predicting peaks in advance can support decision-making in various domains. For example, predicting demand for healthcare services helps adjust nurses’ work schedules in a hospital. This paper proposes a time series mining task named peak prediction, from which most applications that manage resources based on expected demand can benefit. We investigate various approaches, such as conventional machine learning and deep learning methods, to build global and individual models for peak prediction in time series. We evaluate these approaches in a load demand problem on smart meter energy consumption. In this task, a model predicts the customer’s maximum daily energy consumption in the following days (7 days ahead) and identifies which day it will occur. Accurate peak demand predictions for the consumers in a region help utility companies make informed decisions regarding capacity planning, load balancing, and integrating renewable energy sources. Since we are interested in two target variables simultaneously (i.e., peak magnitude and peak position), we evaluated whether multi-target approaches improve single-target models’ performance by correlating both targets in the learning process. Our experimental evaluation considers a dataset comprising three years of daily energy consumption from 1,757 customers. The results show the potential of individual machine learning models induced by simple algorithms such as Linear Regression and XGBoost and multi-target methods’ lack of contribution (or even negative impact) for the performance.

Seismic monitoring systems sift through seismograms in real-time, searching for target events, such as underground explosions. In this monitoring system, a burst of aftershocks (minor earthquakes occur after a major earthquake over days or even years) can be a source of confounding signals. Such a burst of aftershock signals can overload the human analysts of the monitoring system. To alleviate this burden at the onset of a sequence of events (e.g., aftershocks), a human analyst can label the first few of these events and start an online classifier to filter out subsequent aftershock events. We propose an online few-shot classification model FewSig for time series data for the above use case. The framework of FewSig consists of a selective model to identify the high-confidence positive events which are used for updating the models and a general classifier to label the remaining events. Our specific technique uses a %two-level decision tree selective model based on sliding DTW distance and a general classifier model based on distance metric learning with Neighborhood Component Analysis (NCA). The algorithm demonstrates surprising robustness when tested on univariate datasets from the UEA/UCR archive. Furthermore, we show two real-world earthquake events where the FewSig reduces the human effort in monitoring applications by filtering out the aftershock events.

Classification of time series data is an essential task in applications of different fields, from Biology to Sports. Examples include human activity recognition, earthquake detection, insect species identification, and heartbeat analysis. The success of classification algorithms on these problems depends on having varied examples and a reasonable amount of labeled data for training an accurate predictive model. However, obtaining such data can be expensive or impossible in some scenarios, mainly when we collect data at a high sampling rate (e.g., 100 Hz), which is usual for sensors producing time series data. In this paper, we face the challenge of labeled examples scarcity for time series classification by employing data augmentation techniques to increase the number of training examples without collecting and labeling new data. Our method TS-DENSE is a data augmentation strategy based on subclass clustering and simple example modifications. In an evaluation with 48 benchmark datasets, TS-DENSE proved to be a consistent strategy improving the results in most of the problems and settings of the k-Nearest Neighbor algorithm, a state-of-the-art similarity-based method.

The depth of a seismic event is an essential feature to discriminate natural earthquakes from events induced or created by humans. However, estimating the depth of a seismic event with a sparse set of seismic stations is a daunting task, and there is no globally usable method. This paper focuses on developing a machine learning model to accurately estimate the depth of arbitrary seismic events directly from seismograms. Our proposed deep learning architecture is not-so-deep compared to commonly found models in the literature for related tasks, consisting of two loosely connected levels of neural networks, associated with the seismic stations at the higher level and the individual channels of a station at the lower level. Thus, the model has significant advantages, including a reduced number of parameters for tuning and better interpretability to geophysicists. We evaluate our solution on seismic data collected from the SCEDC (Southern California Earthquake Data Center) catalog for regional events in California. The model can learn waveform features specific to a set of stations, while it struggles to generalize to completely novel sets of event sources and stations. In a simplified setting of separating shallow events from deep ones, the model achieved an 86.5% F1-score using the Southern California stations.

Monitoring systems have hundreds or thousands of distributed sensors gathering and transmitting real-time streaming data. The early detection of events in these systems, such as an earthquake in a seismic monitoring system, is the base for essential tasks as warning generations. To detect such events is usual to compute pairwise correlation across the disparate signals generated by the sensors. Since the data sources (e.g., sensors) are spatially separated, it is essential to consider the lagged correlation between the signals. Besides, many applications require to process a specific band of frequencies depending on the event’s type, demanding a pre-processing step of filtering before computing correlations. Due to the high speed of data generation and a large number of sensors in these systems, the operations of filtering and lagged cross-correlation need to be efficient to provide real-time responses without data losses. This article proposes a technique named FilCorr that efficiently computes both operations in one single step. We achieve an order of magnitude speedup by maintaining frequency transforms over sliding windows. Our method is exact, devoid of sensitive parameters, and easily parallelizable. Besides our algorithm, we also provide a publicly available real-time system named Seisviz that employs FilCorr in its core mechanism for monitoring a seismometer network. We demonstrate that our technique is suitable for several monitoring applications as seismic signal monitoring, motion monitoring, and neural activity monitoring.

This paper introduces a new pattern mining task that considers aligning or joining a set of time series based on an arbitrary number of subsequences (i.e., patterns) with arbitrary lengths. Joining multiple time series along common patterns can be pivotal in clustering and summarizing large time series datasets. An exact algorithm to join hundreds of time series based on multi-length patterns is impractical due to the high computational costs. This paper proposes a fast algorithm named MultiPAL to join multiple time series at interactive speed to summarize large time series datasets. The algorithm exploits Matrix Profiles of the individual time series to enable a greedy search over possible joins. The algorithm is orders of magnitude faster than the exact solution and can utilize hundreds of Matrix Profiles. We evaluate our algorithm for sequential mining on data from various real-world domains, including power management and bioacoustics monitoring.

Changes in data distribution of streaming data (i.e., concept drifts), constitute a central issue in online data mining. The main reason is that these changes are responsible for outdating stream learning models, reducing their predictive performance over time. A common approach adopted by real-time adaptive systems to deal with concept drifts is to employ detectors that indicate the best time for updates. However, an unrealistic assumption of most detectors is that the labels become available immediately after data arrives. In this paper, we introduce an unsupervised and model-independent concept drift detector suitable for high-speed and high-dimensional data streams in realistic scenarios with the scarcity of labels. We propose a straightforward two-dimensional representation of the data aiming faster processing for detection. We develop a simple adaptive drift detector on this visual representation that is efficient for fast streams with thousands of features and is accurate as existing costly methods that perform various statistical tests. Our method achieves better performance measured by execution time and accuracy in classification problems for different types of drifts, including abrupt, oscillating, and incremental. Experimental evaluation demonstrates the versatility of the method in several domains, including astronomy, entomology, public health, political science, and medical science.

An essential task on streaming time series data is to compute pairwise correlation across disparate signal sources to identify significant events. In many monitoring applications, such as geospatial monitoring, motion monitoring and critical infrastructure monitoring, correlation is observed at various frequency bands and temporal lags. In this paper, we consider computing filtered and lagged correlation on streaming time series data, which is challenging because the computation must be “in-sync” with the incoming stream for any detected events to be useful. We propose a technique to compute filtered and lagged correlation on streaming data efficiently by merging two individual operations: filtering and cross-correlations. We achieve an order of magnitude speed-up by maintaining frequency transforms over sliding windows. Our method is exact, devoid of sensitive parameters, and easily parallelizable. We demonstrate our technique in a seismic signal monitoring application.

Machine Learning (ML) is becoming a ubiquitous technology employed in many real-world applications. In some applications, sensors measure the environment while ML algorithms are responsible for interpreting the data. These systems often face three main restrictions: power consumption, cost, and lack of infrastructure. Therefore, we need highly-efficient classifiers suitable to execute in unresourceful hardware. However, this scenario conflicts to the state-of-practice of ML, in which classifiers are frequently implemented in high-level interpreted languages, make unrestricted use of floating-point operations and assume plenty of resources. In this paper, we present a software tool named EmbML that implements a pipeline to develop classifiers for low-powered embedded systems. It starts with learning a classifier using popular software packages or libraries. Then, EmbML converts the classifier into a carefully crafted C++ code with support for embedded hardware. Our experimental evaluation shows that EmbML classifiers present competitive results in terms of accuracy, time and memory cost.

The choice of the most promising algorithm to model and predict a particular phenomenon is one of the most prominent activities of the temporal data forecasting. Forecasting (or prediction), similarly to other data mining tasks, uses empirical evidence to select the most suitable model for a problem at hand since no modeling method can be considered as the best. However, according to our systematic literature review of the last decade, few scientific publications rigorously expose the benefits and limitations of the most popular algorithms for time series prediction. At the same time, there is a limited performance record of these models when applied to complex and highly nonlinear data. In this paper, we present one of the most extensive, impartial and comprehensible experimental evaluations ever done in the time series prediction field. From 95 datasets, we evaluate eleven predictors, seven parametric and four non-parametric, employing two multi-step-ahead projection strategies and four performance evaluation measures. We report many lessons learned and recommendations concerning the advantages, drawbacks, and the best conditions for the use of each model. The results show that SARIMA is the only statistical method able to outperform, but without a statistical difference, the following machine learning algorithms: ANN, SVM, and kNN-TSPI. However, such forecasting accuracy comes at the expense of a larger number of parameters. The evaluated datasets, as well detailed results achieved by different indexes as MSE, Theil’s U coefficient, POCID, and a recently-proposed multi-criteria performance measure are available online in our repository. Such repository is another contribution of this paper since other researchers can replicate our results and evaluate their methods more rigorously. The findings of this study will impact further research on this topic since they provide a broad insight into models selection, parameters setting, evaluation measures, and experimental setup.

Many real-world applications, such as those related to sensors, allow collecting large amounts of inexpensive unlabeled sequential data. However, the use of supervised machine learning methods is frequently hindered by the high costs involved in gath

The majority of evolving data streams classification algorithms assume that the actual labels of the predicted examples are readily available without any time delay just after a prediction is made. However, given the high label costs, dependence of an expert, limitations in data transmission or even restrictions imposed by the problem's nature, there is a large number of real-world applications in which the availability of actual labels is infinitely delayed (never available). In these cases, it is necessary the use of algorithms that does not follow the traditional process of monitoring the error rate to detect changes in data distribution and uses the most recent labeled data to update the classification model. In this paper, we propose the method MClassification to classify evolving data streams with infinitely delayed labels. Our method is inspired on the use of Micro-Cluster representation from online clustering algorithms. Considering the presence of incremental drifts, our approach uses a distance-based strategy to maintain the Micro-Clusters' positions updated. An evaluation in several synthetic and real data shows that MClassification achieves competitive accuracy results to state-of-the-art methods and adequate computational cost. The main advantage of the proposed method is the absence of critical parameters that require user's prior knowledge, as occurs with rival methods.

Throughout the history, insects had been intimately connected to humanity, in both positive and negative ways. Insects play an important part in crop pollination, on the other hand, some of them spread diseases that kill millions of people every year. Effective control of harmful insects while having little impact to beneficial insects and environment is extremely important. Recently, an intelligent trap that uses laser sensors was presented to control the population of target insects. The device could record and analyze sensor signals when an insect passes through the trap and make quick decisions whether to catch it or not. The effectiveness of the trap relies on the correct choice of classification algorithm to perform the insect detection. In this paper, we propose to use a deep neural network with maximum correntropy criterion (MCC) for reliable classification of insects in real-time. Experimental results show that, deep networks are effective for learning stable features from brief insect passage signals. By replacing the mean square error cost with MCC, the robustness of autoencoders against noise is improved significantly and robust features could be learned. The method is tested on five species of insects and a total of 5325 passages. High classification accuracy of 92.1% is achieved. Compared with previously applied methods, better classification performance is obtained using only 10% of the computation time. Therefore, our method is efficient and reliable for online insect detection.

Graph-based semi-supervised learning (SSL) algorithms perform well when the data lie on a low-dimensional manifold. Although these methods achieved satisfactory performance on a variety of domains, they have not been effectively evaluated on time series classification. In this paper, we provide a comprehensive empirical comparison of state-of-the-art graph-based SSL algorithms combined with a variety of graph construction methods in order to evaluate them on time series transductive classification tasks. Through a detailed experimental analysis using recently proposed empirical evaluation models, we show strong and weak points of these classifiers concerning both performance and stability with respect to graph construction and parameter selection. Our results show that some hypotheses raised on previous work do not hold in the time series domain while others may only hold under mild conditions.

Automatic classification of musical instruments is an important task for music transcription as well as for professionals such as audio designers, engineers and musicians. Unfortunately, only a limited amount of effort has been conducted to automatically classify percussion instrument in the last years. The studies that deal with percussion sounds are usually restricted to distinguish among the instruments in the drum kit such as toms vs. snare drum vs. bass drum vs. cymbals. In this paper, we are interested in a more challenging task of discriminating sounds produced by the same percussion instrument. Specifically, sounds from different drums cymbals types. Cymbals are known to have indefinite pitch, nonlinear and chaotic behavior. We also identify how the sound of a specific cymbal was produced (e.g., roll or choke movements performed by a drummer). We achieve an accuracy of 96.59% for cymbal type classification and 91.54% in a classification problem with 12 classes which represent the cymbal type and the manner or region that the cymbals are struck. Both results were obtained with Support Vector Machine algorithm using the Line Spectral Frequencies as audio descriptor. We believe that our results can be useful for a more detailed automatic drum transcription and for other related applications as well for audio professionals.

Data stream classification algorithms for nonstationary environments frequently assume the availability of class labels, instantly or with some lag after the classification. However, certain applications, mainly those related to sensors and robotics, involve high costs to obtain new labels during the classification phase. Such a scenario in which the actual labels of processed data are never available is called extreme verification latency. Extreme verification latency requires new classification methods capable of adapting to possible changes over time without external supervision. This paper presents a fast, simple, intuitive and accurate algorithm to classify nonstationary data streams in an extreme verification latency scenario, namely Stream Classification Algorithm Guided by Clustering – SCARGC. Our method consists of a clustering followed by a classification step applied repeatedly in a closed loop fashion. We show in several classification tasks evaluated in synthetic and real data that our method is faster and more accurate than the state-of-the-art.

Graph-based semi-supervised learning (SSL) algorithms perform well on a variety of domains, such as digit recognition and text classification, when the data lie on a low-dimensional manifold. However, it is surprising that these methods have not been effectively applied on time series classification tasks. In this paper, we provide a comprehensive empirical comparison of state-of-the-art graph-based SSL algorithms with respect to graph construction and parameter selection. Specifically, we focus in this paper on the problem of time series transductive classification on imbalanced data sets. Through a comprehensive analysis using recently proposed empirical evaluation models, we confirm some of the hypotheses raised on previous work and show that some of them may not hold in the time series domain. From our results, we suggest the use of the Gaussian Fields and Harmonic Functions algorithm with the mutual k-nearest neighbors graph weighted by the RBF kernel, setting k = 20 on general tasks of time series transductive classification on imbalanced data sets.

Time series are present in many pattern recognition applications related to medicine, biology, astronomy, economy, and others. In particular, the classification task has attracted much attention from a large number of researchers. In such a task, empirical researches has shown that the 1-Nearest Neighbor rule with a distance measure in time domain usually performs well in a variety of application domains. However, certain time series features are not evident in time domain. A classical example is the classification of sound, in which representative features are usually present in the frequency domain. For these applications, an alternative representation is necessary. In this work we investigate the use of recurrence plots as data representation for time series classification. This representation has well-defined visual texture patterns and their graphical nature exposes hidden patterns and structural changes in data. Therefore, we propose a method capable of extracting texture features from this graphical representation, and use those features to classify time series data. We use traditional methods such as Grey Level Co-occurrence Matrix and Local Binary Patterns, which have shown good results in texture classification. In a comprehensible experimental evaluation, we show that our method outperforms the state-of-the-art methods for time series classification.

There is a huge increase of interest for time series methods and techniques. Virtually every piece of information collected from human, natural, and biological processes is susceptible to changes over time, and the study of how these changes occur is a central issue in fully understanding such processes. Among all time series mining tasks, classification is likely to be the most prominent one. In time series classification there is a significant body of empirical research that indicates that k-nearest neighbor rule in the time domain is very effective. However, certain time series features are not easily identified in this domain and a change in representation may reveal some significant and unknown features. In this work, we propose the use of recurrence plots as representation domain for time series classification. Our approach measures the similarity between recurrence plots using Campana-Keogh (CK-1) distance, a Kolmogorov complexity-based distance that uses video compression algorithms to estimate image similarity. We show that recurrence plots allied to CK-1 distance lead to significant improvements in accuracy rates compared to Euclidean distance and Dynamic Time Warping in several data sets. Although recurrence plots cannot provide the best accuracy rates for all data sets, we demonstrate that we can predict ahead of time that our method will outperform the time representation with Euclidean and Dynamic Time Warping distances.

Throughout the history, insects have had an intimate relationship with humanity, both positive and negative. Insects are vectors of diseases that kill millions of people every year and, at the same time, insects pollinate most of the world's food production. Consequently, there is a demand for new devices able to control the populations of harmful insects while having a minimal impact on beneficial insects. In this paper, we present an intelligent trap that uses a laser sensor to selectively classify and catch insects. We perform an extensive evaluation of different feature sets from audio analysis and machine learning algorithms to construct accurate classifiers for the insect classification task. Support Vector Machines achieved the best results with a MFCC feature set, which consists of coefficients from frequencies scaled according to the human auditory system. We evaluate our classifiers in multiclass and binary class settings, and show that a binary class classifier that recognizes the mosquito species achieved almost perfect accuracy, assuring the applicability of the proposed intelligent trap.