Training large neural networks and merging task-specific models both exploit low-rank structure and require parameter importance estimation, yet these challenges have been pursued in isolation. Current workflows compute curvature information during training, discard it, then recompute similar information for merging—wasting computation and discarding valuable trajectory data. We introduce a unified framework that maintains factorized momentum and curvature statistics during training, then reuses this information for geometry-aware model composition. The proposed method incurs modest memory overhead (approximately 30% over AdamW) to accumulate task saliency scores that enable curvature-aware merging. These scores, computed as a byproduct of optimization, provide importance estimates comparable to post-hoc Fisher computation while producing merge-ready models directly from training. We establish convergence guarantees for non-convex objectives with approximation error bounded by gradient singular value decay. On natural language understanding benchmarks, curvature-aware parameter selection outperforms magnitude-only baselines across all sparsity levels, with multi-task merging improving 1.6% over strong baselines. The proposed framework exhibits rank-invariant convergence and superior hyperparameter robustness compared to existing low-rank optimizers. By treating the optimization trajectory as a reusable asset rather than discarding it, our approach demonstrates that training-time curvature information suffices for effective model composition, enabling a unified training-merging pipeline.

Abstract Predicting the occurrence of crop pests is becoming a crucial task in modern agriculture to facilitate farmers’ decision-making. One of the most significant pests is the olive fruit fly, a public concern because it causes damage that compromises oil quality, increasing acidity and altering its flavor. This paper proposes a hybrid deep learning model to predict the presence of olive flies in crops. This model is based on an autoencoder and an automated deep feed-forward neural network. First, the autoencoder neural network learns a representation of the data and then the automated deep feed-forward neural network automatically determines the best values for the hyperparameters in order to obtain the prediction of the number of flies caught in traps from the dataset generated by the autoencoder. On the other hand, farmers to trust the proposed deep learning models need these models to be explainable. Thus, explainable artificial intelligence techniques are applied to the produced models to interpret the results. Results using a dataset from different sources such as satellite image band data, vegetation indices, and meteorological variables are reported. The performance of the proposed model has been compared with classical benchmark algorithms and a deep learning model recently published in the literature. In addition, the comparison includes the automated deep feed-forward neural network individually to show how the autoencoder network improves the accuracy of predictions.

Abstract Electricity market forecasting is very useful for the different actors involved in the energy sector to plan both the supply chain and market operation. Nowadays, energy demand data are data coming from smart meters and have to be processed in real-time for more efficient demand management. In addition, electricity prices data can present changes over time such as new patterns and new trends. Therefore, real-time forecasting algorithms for both demand and prices have to adapt and adjust to online data in order to provide timely and accurate responses. This work presents a new algorithm for electricity demand and prices forecasting in real-time. The proposed algorithm generates a prediction model based on the k-nearest neighbors algorithm, which is incrementally updated in an online scenario considering both changes to existing patterns and adding new detected patterns to the model. Both time-frequency and error threshold based model updates have been evaluated. Results using energy demand from 2007 to 2016 and prices data for different time periods from the Spanish electricity market are reported and compared with other benchmark algorithms.

This paper introduces a new, model-independent, metric, called RExQUAL, for quantifying the quality of explanations provided by attribution-based explainable artificial intelligence techniques and compare them. The underlying idea is based on feature attribution, using a subset of the ranking of the attributes highlighted by a model-agnostic explainable method in a forecasting task. Then, association rules are generated using these key attributes as input data. Novel metrics, including global support and confidence, are proposed to assess the joint quality of generated rules. Finally, the quality of the explanations is calculated based on a wise and comprehensive combination of the association rules global metrics. The proposed method integrates local explanations through attribution-based approaches for evaluation and feature selection with global explanations for the entire dataset. This paper rigorously evaluates the new metric by comparing three explainability techniques: the widely used SHAP and LIME, and the novel methodology RULEx. The experimental design includes predicting time series of different natures, including univariate and multivariate, through deep learning models. The results underscore the efficacy and versatility of the proposed methodology as a quantitative framework for evaluating and comparing explainable techniques.

Water consumption forecasting is an essential tool for water management, as it allows for efficient planning and allocation of water resources, an undervalued but indispensable resource for all living beings. With the increasing demand for accurate and timely water forecasting, traditional forecasting methods are proving to be insufficient. Deep learning techniques, which have shown remarkable performance in a wide range of applications, offer a promising approach to address the challenges of water consumption forecasting. In this work, the use of deep learning models for medium-term water consumption forecasting of residential areas is explored. A deep feed-forward neural network is developed to predict water consumption of a company’s customers for the next quarter. First, customers are grouped according to their consumption as these customers include both household consumers and special consumers such as public swimming pools, sports halls or small industries. Then, a deep feed-forward neural network is designed for household customers by obtaining the optimal values for those hyperparameters that have a great influence on the network performance. Results are reported using a real-world dataset composed of the water consumption from 1999 to 2015 on a quarterly basis, corresponding to 3262 clients of a water supply company. Finally, the proposed algorithm is evaluated by comparing it with other reference algorithms including an LSTM network.

Several interrelated variables typically characterize real-world processes, and a time series cannot be predicted without considering the influence that other time series might have on the target time series. This work proposes a novel algorithm to forecast multivariate big data time series. This new general-purpose approach consists first of a previous pattern recognition performed jointly using all time series that form the multivariate time series and then predicts the target time series by searching for similarities between pattern sequences. The proposed algorithm is designed to tackle multivariate time series forecasting problems within the context of big data. In particular, the algorithm has been developed with a distributed nature to enhance its efficiency in analyzing and processing large volumes of data. Moreover, the algorithm is straightforward to use, with only two parameters needing adjustment. Another advantage of the MV-bigPSF algorithm is its ability to perform multi-step forecasting, which is particularly useful in many practical applications. To evaluate the algorithm’s performance, real-world data from Uruguay’s power consumption has been utilized. Specifically, MV-bigPSF has been compared with both univariate and multivariate methods. Regarding the univariate ones, MV-bigPSF improved 12.8% in MAPE compared to the second-best method. Regarding the multivariate comparison, MV-bigPSF improved 44.8% in MAPE with respect to the second most accurate method. Regarding efficiency, the execution time of MV-bigPSF was 1.83 times faster than the second-fastest multivariate method, both in a single-core environment. Therefore, the proposed algorithm can be a valuable tool for practitioners and researchers working in multivariate time series forecasting, particularly in big data applications.

Time series data can be defined as a chronological sequence of observations on a variable of interest. A streaming time series is a time series that arrives continuously at high speed and has a data distribution that may change over time. Streaming time series data usually comes from electronic devices such as sensors and many of the applications dealing with streaming data in Industry 4.0 require real-time responses. Performing real-time forecasting offers the possibility to consider new types of patterns in the incoming streaming data, which is not possible when working with batch models. This paper presents a new approach to detect novelties and anomalies in real-time using a nearest-neighbors based forecasting algorithm. The algorithm works with an offline base model that is updated as stream data arrives following an incremental learning approach. It detects unknown patterns called novelties and anomalies. Novelties are included in the model in an online way and anomalies trigger an alarm as they present unexpected behaviors that need to be specifically analyzed. The algorithm has been tested with Spanish electricity demand data. Results show that the prediction errors obtained when the model is updated considering novelties and anomalies are lower than the errors obtained when the model is not updated. Thus, the model adjusts in real-time to the new patterns of data providing accurate errors and real-time predictions.

Machine and deep learning has become one of the most useful tools in the last years as a diagnosis-decision-support tool in the health area. However, it is widely known that artificial intelligence models are considered a black box and most experts experience difficulties explaining and interpreting the models and their results. In this context, explainable artificial intelligence is emerging with the aim of providing black-box models with sufficient interpretability so that models can be easily understood and further applied. Obstructive sleep apnea is a common chronic respiratory disease related to sleep. Its diagnosis nowadays is done by processing different data signals, such as electrocardiogram or respiratory rate. The waveform of the respiratory signal is of importance too. Machine learning models could be applied to the signal's analysis. Data from a polysomnography study for automatic sleep apnea detection have been used to evaluate the use of the Local Interpretable Model-Agnostic (LIME) library for explaining the health data models. Results obtained help to understand how several features have been used in the model and their influence in the quality of sleep.

Precision agriculture focuses on the development of site-specific harvest considering the variability of each crop area. Vegetation indices allow the study and delineation of different characteristics of each field zone, generally invisible to the naked-eye. This paper introduces a new big data triclustering approach based on evolutionary algorithms. The algorithm shows its capability to discover three-dimensional patterns on the basis of vegetation indices from vine crops. Different vegetation indices have been tested to find different patterns in the crops. The results reported using a vineyard crop located in Portugal depicts four areas with different moisture stress particularities that can lead to changes in the management of the vineyard. Furthermore, scalability studies have been performed, showing that the proposed algorithm is suitable for dealing with big datasets.

Time series forecasting has become indispensable for multiple applications and industrial processes. Currently, a large number of algorithms have been developed to forecast time series, all of which are suitable depending on the characteristics and patterns to be inferred in each case. In this work, a new algorithm is proposed to predict both univariate and multivariate time series based on a combination of clustering, classification and forecasting techniques. The main goal of the proposed algorithm is first to group windows of time series values with similar patterns by applying a clustering process. Then, a specific forecasting model for each pattern is built and training is only conducted with the time windows corresponding to that pattern. The new algorithm has been designed using a flexible framework that allows the model to be generated using any combination of approaches within multiple machine learning techniques. To evaluate the model, several experiments are carried out using different configurations of the clustering, classification and forecasting methods that the model consists of. The results are analyzed and compared to classical prediction models, such as autoregressive, integrated, moving average and Holt-Winters models, to very recent forecasting methods, including deep, long short-term memory neural networks, and to well-known methods in the literature, such as k nearest neighbors, classification and regression trees, as well as random forest.

Nowadays, electricity is a basic commodity necessary for the well-being of any modern society. Due to the growth in electricity consumption in recent years, mainly in large cities, electricity forecasting is key to the management of an efficient, sustainable and safe smart grid for the consumer. In this work, a deep neural network is proposed to address the electricity consumption forecasting in the short-term, namely, a long short-term memory (LSTM) network due to its ability to deal with sequential data such as time-series data. First, the optimal values for certain hyper-parameters have been obtained by a random search and a metaheuristic, called coronavirus optimization algorithm (CVOA), based on the propagation of the SARS-Cov-2 virus. Then, the optimal LSTM has been applied to predict the electricity demand with 4-h forecast horizon. Results using Spanish electricity data during nine years and half measured with 10-min frequency are presented and discussed. Finally, the performance of the proposed LSTM using random search and the LSTM using CVOA is compared, on the one hand, with that of recently published deep neural networks (such as a deep feed-forward neural network optimized with a grid search) and temporal fusion transformers optimized with a sampling algorithm, and, on the other hand, with traditional machine learning techniques, such as a linear regression, decision trees and tree-based ensemble techniques (gradient-boosted trees and random forest), achieving the smallest prediction error below 1.5%.

Time-series clustering is the process of grouping time series with respect to their similarity or characteristics. Previous approaches usually combine a specific distance measure for time series and a standard clustering method. However, these approaches do not take the similarity of the different subsequences of each time series into account, which can be used to better compare the time-series objects of the dataset. In this article, we propose a novel technique of time-series clustering consisting of two clustering stages. In a first step, a least-squares polynomial segmentation procedure is applied to each time series, which is based on a growing window technique that returns different-length segments. Then, all of the segments are projected into the same dimensional space, based on the coefficients of the model that approximates the segment and a set of statistical features. After mapping, a first hierarchical clustering phase is applied to all mapped segments, returning groups of segments for each time series. These clusters are used to represent all time series in the same dimensional space, after defining another specific mapping process. In a second and final clustering stage, all the time-series objects are grouped. We consider internal clustering quality to automatically adjust the main parameter of the algorithm, which is an error threshold for the segmentation. The results obtained on 84 datasets from the UCR Time Series Classification Archive have been compared against three state-of-the-art methods, showing that the performance of this methodology is very promising, especially on larger datasets.

Abstract In this paper, we consider the task of predicting the electricity power generated by photovoltaic solar systems for the next day at half‐hourly intervals. We introduce DL, a deep learning approach based on feed‐forward neural networks for big data time series, which decomposes the forecasting problem into several sub‐problems. We conduct a comprehensive evaluation using 2 years of Australian solar data, evaluating accuracy and training time, and comparing the performance of DL with two other advanced methods based on neural networks and pattern sequence similarity. We investigate the use of multiple data sources (solar power and weather data for the previous days, and weather forecast for the next day) and also study the effect of different historical window sizes. The results show that DL produces competitive accuracy results and scales well, and is thus a highly suitable method for big data environments.

Accurate forecasting of the power generated by PhotoVoltaic (PV) systems is needed for the successful integration of solar power into the electricity grid. In this paper, we consider the task of simultaneously forecasting the PV power output for the next day at half-hourly intervals using only previous PV power data. We propose a number of methods for constructing static and dynamic ensembles of neural networks. The static ensembles are based on random sampling and random feature selection, and the dynamic ones adaptively weight the contribution of the ensemble members based on their recent performance. We conduct an evaluation using Australian solar PV data for two years, and compare the results with state-of-the-art single prediction models and classical ensemble models. The results show that the proposed static ensembles are beneficial, achieving higher accuracy than the single neural networks and the other single and ensemble models used for comparison. The dynamic ensemble versions further improve the accuracy, demonstrating the potential of adaptively combining the predictions of ensemble members for accurate solar power forecasting.

We present INN, a new approach for predicting the hourly electricity load profile for the next day from a time series of previous electricity loads. It uses an iterative methodology to make the predictions for the 24-hour forecasting horizon. INN combines an efficient mutual information feature selection method with a neural network forecasting algorithm. We evaluate INN using two years of electricity load data for Australia, Portugal and Spain. The results show that it provides accurate predictions, outperforming three state-of-the-art approaches (weighted nearest neighbor, pattern sequence similarity and iterative linear regression), and a number of baselines. INN is also more accurate and efficient than a non-iterative version of the approach. We also found that although the range of load values for the three countries is very different, the load curves show similar patterns, which resulted in more than 90% overlap in the selected lag variables.

We present PSF-NN, a new approach for time series forecasting. It combines prediction based on sequence similarity with neural networks. PSF-NN first generates predictions using the PSF algorithm that are then refined by the neural network component, which also utilizes additional features. We evaluate the performance of PSF-NN using a time series of hourly electricity demands for the state of New South Wales in Australia for three years. The task is to predict an interval of future values simultaneously, i.e. the 24 demands for the next day, instead of predicting just a single future demand. The results showed that the combined PSF-NN approach provides accurate predictions, outperforming the original PSF algorithm and a number of baselines.

The identification of regulatory modules is one of the most important tasks in order to discover disease markers. This paper presents a methodology to infer coexpression networks based on local patterns in gene expression data matrix. In the proposed algorithm two steps can clearly be differentiated. Firstly, a Biclustering procedure that uses a Scatter Search schema to find biclusters and, secondly, a network extraction procedure based on linear correlations among the genes of the previously obtained bicluster. Experimental results from Yeast cell Cycle are reported where three different algorithms have been applied. Also, a possible understanding of one of the obtained networks has been presented from a biological point of view.

This paper presents a new approach to forecast the behavior of time series based on similarity of pattern sequences. First, clustering techniques are used with the aim of grouping and labeling the samples from a data set. Thus, the prediction of a data point is provided as follows: first, the pattern sequence prior to the day to be predicted is extracted. Then, this sequence is searched in the historical data and the prediction is calculated by averaging all the samples immediately after the matched sequence. The main novelty is that only the labels associated with each pattern are considered to forecast the future behavior of the time series, avoiding the use of real values of the time series until the last step of the prediction process. Results from several energy time series are reported and the performance of the proposed method is compared to that of recently published techniques showing a remarkable improvement in the prediction.

In this paper a hybrid metaheuristic for biclustering based on Scatter Search and Genetic Algorithms is presented. A general scheme of Scatter Search has been used to obtain high-quality biclusters, but a way of generating the initial population and a method of combination based on Genetic Algorithms have been chosen. Moreover, in the own algorithm the overlapping among biclusters is controlled adding a penalization term in the fitness function. Experimental results from yeast cell cycle are reported. Finally, the performance of the proposed hybrid algorithm is compared with a genetic algorithm recently published.

A new approach is presented in this work with the aim of predicting time series behaviors. A previous labeling of the samples is obtained utilizing clustering techniques and the forecasting is applied using the information provided by the clustering. Thus, the whole data set is discretized with the labels assigned to each data point and the main novelty is that only these labels are used to predict the future behavior of the time series, avoiding using the real values of the time series until the process ends. The results returned by the algorithm, however, are not labels but the nominal value of the point that is required to be predicted. The algorithm based on labeled (LBF) has been tested in several energy-related time series and a notable improvement in the prediction has been achieved.

Breast cancer is one of the diseases causing the largest number of deaths among women. Its early detection has been proved to be the most effective way to combat it. This work is focused on developing an integral tool able to detect microcalcifications in mammographies, since the presence of these particles is a clear symptom of an incipient cancer. The proposed approach combines two techniques successfully used in other areas separately, such as linear pixel prediction and support-vector machines, in order to obtain almost perfect prediction accuracy. Moreover, a filter has been designed with the aim of decrease the processing time. The result verges on 96% of hits, improving previous works by 6%, on average.