The global energy sector is undergoing a profound and multidimensional transformation driven by decarbonization policies, increasing electrification, large-scale integration of renewable energy sources, and the growing digitalization of energy infrastructures [...]

Effectively reducing the size of the training set for classifier training is a rapidly advancing area of machine learning research. In this context, meta instance selection has emerged as a promising approach. This method transforms the instance selection problem into a classification task and employs a meta-classifier to determine which samples should be kept and which should be removed. In this study, we focus on identifying the optimal meta-classifier by evaluating several popular predictive models, including balanced random forests, standard random forests, gradient-boosted trees, k-nearest neighbors, and multi-layer perceptron networks. The results demonstrate that the balanced random forest outperforms all other models, yielding the best performance by a significant margin.

Data pruning, or instance selection, is an important problem in machine learning especially in terms of nearest neighbour classifier. However, in data pruning which speeds up the prediction phase, there is an issue related to the speed and efficiency of the process itself. In response, the study proposes an approach involving transforming the instance selection process into a classification task conducted in a unified meta-feature space where each instance can be classified and assigned to either the "to keep" or "to remove" class. This approach requires training an appropriate meta-classifier, which can be developed based on historical instance selection results from other datasets using reference instance selection methods as a labeling tool. This work proposes constructing the meta-feature space based on properties extracted from the nearest neighbor graph. Experiments conducted on 17 datasets of varying sizes and five reference instance selection methods (ENN, Drop3, ICF, HMN-EI, and CCIS) demonstrate that the proposed solution achieves results comparable to reference instance selection methods while significantly reducing computational complexity. In the proposed approach, the computational complexity of the system depends only on identifying the k-nearest neighbors for each data sample and running the meta-classifier. Additionally, the study discusses the choice of meta-classifier, recommending the use of Balanced Random Forest.

This article focuses on the problem of building a real-world predictive maintenance system for hydraulic piston pumps. Particular attention is given to the issue of limited data availability regarding the failure state of systems with a damaged valve plate. The main objective of this work was to analyze the impact of imbalanced data on the quality of the failure prediction system. Several data balancing techniques, including oversampling, undersampling, and combined methods, were evaluated to overcome the limitations. The dataset used for evaluation includes recordings from eleven sensors, such as pressure, flow, and temperature, registered at various points in the hydraulic system. It also includes data from three additional vibration sensors. The experiments were conducted with imbalance ratios ranging from 0.5% to a fully balanced dataset. The results indicate that two methods, Borderline SMOTE and SMOTE+Tomek Links, dominate. These methods allowed the system to achieve the highest performance on a completely new dataset with different levels of damaged valve plates, for the balance rate larger than three percent. Furthermore, for balance rates below one percent, the use of data balancing methods may adversely affect the model. Finally, our results indicate the limitations of the use of cross-validation procedures when assessing data balancing methods.

We present a preliminary study of a model-agnostic method called prototype pair decomposition that generates simple and accurate decision rules from datasets. The research focuses on its application to decision trees. It starts by selecting representative prototypes obtained by a prototype construction method, then pairs of prototypes from opposite classes are determined. These pairs define subspaces containing a fragment of the decision boundary in which a shallow decision tree is applied to extract simple decision rules consisting of a few premises. The results indicate that the proposed solution allows the extraction of locally competent simple rules that are comparable in terms of classification accuracy to a large and complex set of global rules obtained from standard decision trees.

The classification of Unexploded Ordnance (UXO) from magnetometer data is a critical but challenging task, frequently hindered by the data scarcity required for training robust machine learning models. To address this, we leverage a high-fidelity digital twin to generate a comprehensive dataset of magnetometer signals from both UXO and non-UXO objects, incorporating complex remanent magnetization effects. In this study, we design and evaluate a custom Convolutional Neural Network (CNN) for UXO classification and compare it against classical machine learning baseline, including Random Forest and kNN. Our CNN model achieves a balanced accuracy of 84.65%, significantly outperforming traditional models that exhibit performance collapse under slight distortions such as additive noise, drift, and time-wrapping. Additionally, we present a compact two-block CNN variant that retains competitive accuracy while reducing the number of learnable parameters by approximately 33%, making it suitable for real-time onboard classification in underwater vehicle missions. Through extensive ablation studies, we confirm that architectural components, such as residual skip connections and element-wise batch normalization, are crucial for achieving model stability and performance. The results also highlight the practical implications of underwater vehicles for survey design, emphasizing the need to mitigate signal drift and maintain constant survey speeds. This work not only provides a robust deep learning model for UXO classification, but also offers actionable suggestions for improving both model deployment and data acquisition protocols in the field.

Reducing the energy consumption of a data centre while maintaining the requirements of the compute resources is a challenging problem that requires intelligent system design. It even becomes more challenging when dealing with an operating data centre. To achieve that goal without compromising the working conditions of the compute resources, a temperature model is needed that estimates the temperature within the hot corridor of the cooling system based on the properties of the external weather conditions and internal conditions such as server energy consumption, and cooling system state. In this paper, we discuss the dataset creation process as well as the process of evaluating a model for forecasting the temperature in the warm corridor of the data centre. The proposed solution compares two new neural network architectures, namely Time-Series Dense Encoder (TiDE) and Time-Series Mixer (TSMixer) with classical methods such as Random Forest and XGBoost and AutoARIMA. The obtained results indicate that the lowest prediction error was achieved by the TiDE model allowing to achieve 0.1270 of N-RMSE followed by the XGBoost model with 0.1275 of N-RMSE. The additional analysis indicates a limitation of the use of the XGBoost model which tends to underestimate temperature as it approaches higher values, which is particularly important in avoiding safety conditions violations of the compute units.

This study presents two novel approaches for developing a multiple-meta-instance selection method, an advanced algorithm designed for efficient pruning of training sample in classification problems. The proposed meta-instance selection framework reformulates the traditional instance selection problem by introducing a meta-feature space, a problem-agnostic representation space. The transformation enables instance selection to be framed as a classification task in the meta-feature space, facilitating efficient computation with a time complexity of O(nlog(n)). A standard classification algorithm, such as Random Forest, can then be employed in the meta-feature space to determine the inclusion or exclusion of individual samples. To enhance performance, we explore two strategies for combining multiple meta-instance selection algorithms: (1) constructing an ensemble of meta-classifiers and (2) concatenating many meta-sets. Experimental evaluations demonstrate that the meta-set concatenation approach surpasses both classical instance selection techniques and existing meta-instance selection methods. Moreover, the proposed algorithm significantly accelerates the instance selection process—achieving even by two or three orders of magnitude speed-up, depending on dataset size and the reference instance selection method.

The article evaluates the use of machine learning methods to predict valve plate failures in hydraulic pumps. To conduct this assessment, a laboratory setup was developed that allowed the creation of a training dataset containing both the normal operating state of the pump and the operating state with three different valve plate damages. Five different predictive models were evaluated, including three composite models based on decision trees: Random Forest, Gradient-Boosted Trees, and Rotation Forest, as well as the kNN algorithm and a neural network. The analysis showed that neural networks achieved the highest prediction accuracy, enabling a prediction accuracy level of 89%. The kNN algorithm ranked second, and tree-based algorithms performed 4% worse than the neural network. Additionally, the attribute importance analysis revealed that leak flow, pressure output together with output flow, are the most important parameters for accurate prediction.

The article introduces datasets representing piston pump failures along with the experimental evaluation of various machine learning classification models. It starts with a detailed description of three classification datasets consisting of three different levels of valve plate damages and signals recorded from sensors used in classical hydraulic systems (pressure, temperature, flow). The obtained datasets consist of 100k (Failure 1), 30k (Failure 2) and 30k (Failure 3) samples and eight attributes. Then a broad range of classifiers are evaluated including three ensemble models based on decision trees: Random Forest, Gradient-Boosted Trees, and Rotation Forest, as well as the kNN algorithm and a neural network. The analysis showed that neural networks achieved the highest prediction accuracy, enabling a prediction accuracy level of 89%. The kNN algorithm ranked second, and tree-based algorithms performed 4% worse than the neural network. Next, the attribute importance analysis revealed that leak flow, pressure output, pressure of the leak line, and oil temperature are the most important parameters for accurate predictions. Additionally, the research includes a sensitivity analysis of the best classifier to verify the impact of sensor measurements or other noise indicators on the prediction model performance. The analysis indicates a 5% margin of measurement quality.

This research aimed to develop a dataset of acoustic images recorded by a forward-looking sonar mounted on an underwater vehicle, enabling the classification of unexploded ordnances (UXOs) and objects other than unexploded ordnance (nonUXOs). The dataset was obtained using digital twin simulations performed in the Gazebo environment utilizing plugins developed within the DAVE project. It consists of 69,444 sample images of 512 × 399 resolution organized in two classes annotated as UXO and nonUXO. The obtained dataset was then evaluated by state-of-the-art image classification methods using off-the-shelf models and transfer learning techniques. The research included VGG16, ResNet34, ResNet50, ViT, RegNet, and Swin Transformer. Its goal was to define a base rate for the development of other specialized machine learning models. Neural network experiments comprised two stages—pre-training of only the final layers and pre-training of the entire network. The experiments revealed that to obtain high accuracy, it is required to pre-train the entire network, under which condition, all the models achieved comparable performance, reaching 98% balanced accuracy. Surprisingly, the highest accuracy was obtained by the VGG model.

Scanning underwater areas using magnetometers in search of unexploded ordnance is a difficult challenge, where machine learning methods can find a significant application. However, this requires the creation of a dataset enabling the training of prediction models. Such a task is difficult and costly due to the limited availability of relevant data. To address this challenge in the article, we propose the use of numerical modeling to solve this task. The conducted experiments allow us to conclude that it is possible to obtain high compliance with the numerical model based on the finite element method with the results of physical tests. Additionally, the paper discusses the methodology of simplifying the computational model, allowing for an almost three times reduction in the calculation time without affecting model quality. The article also presents and discusses the methodology for generating a dataset for the discrimination of UXO/non-UXO objects. According to that methodology, a dataset is generated and described in detail including assumptions on objects considered as UXO and nonUXO.

Renewable energy sources, such as wind turbines, have become much more prevalent in recent years, and thus a popular form of energy generation. This is in part due to the ‘Fit for 55’ EU initiative, and in part, to rising fossil fuel prices, as well as the perceived requirement for nations to have power independence, and due to the influence of renewable energy sources we can see a marked increase in large wind farms in particular. However, wind farms by their very nature are highly inconsistent regarding power generation and are weather-dependent, thus presenting several challenges for transmission system operators. One of the options to overcome these issues is a system being able to forecast the generated power in a wide-ranging period—ranging from 15 min up to 36 h, and with an adequate resolution. Such a system would better help manage the power grid and allow for greater utilization of the green energy produced. In this document, we present a process of development for such a system, along with a comparison of the various steps of the process, including data preparation, feature importance analysis, and the impact of various data sources on the forecast horizon. Lastly, we also compare multiple machine learning models and their influence on the system quality and execution time. Additionally, we propose an ensemble that concatenates predictions over the forecast horizon. The conducted experiments have been evaluated on seven wind farms located in Central Europe. Out of the experiments conducted, the most efficient solution with the lowest error rate and required computational resources has been obtained for random forest regression, and two independent models; one for the short-term horizon, and the other, for the mid- to long-term horizon, which was combined into one forecasting system.

Unexploded ordnance (UXO) dumped in water reservoirs pose a serious environmental and human safety hazard. Various ways of economically solving this problem are being sought. One of them is the use of machine learning methods for the automatic classification of dangerous objects based on the recorded signals. The paper presents the preliminary results on the use of machine learning methods applied to raw magnetometry data generated in a virtual environment based on the concept of a digital twin. This introduces a different approach to a standard approach, which is based on the inverse problem, where the signals are mapped to the magnetic dipole model. Conducted research points out that the highest performance can be obtained with neural networks, and a direct classification based on the raw signals allows to achieve accuracy of up to 93

Data selection, which includes feature and instance selection, is often an important step in building prediction systems. Genetic algorithms (GA) frequently allow finding better solutions than classical methods in many areas. This is also true for the instance selection task. The main difficulties and challenges in GA-based instance selection are high computational complexity and decreasing performance with the dataset size growth. This is caused by the fact that each instance is encoded in one chromosome position. Hence bigger datasets result in longer chromosomes. The main contribution of this paper addresses the above problems in a three-step process. In the first step the dataset is divided into several consistent regions by fuzzy clustering. Then GA-based instance selection is performed independently within each cluster. Finally ensemble voting provides seamless aggregation of the partial results from the overlapping clusters. This improves dataset exploitation by more localized search and also takes the advantage of ensemble methods. This method significantly improves the predictive model performance and data reduction in comparison to instance selection performed on the whole training dataset.

The overwhelming majority of methodologies for the flexible flow shop scheduling problem proposed so far have a common feature, which is the assumption of constant time and cost for the execution of individual technological operations (ignoring an optimal selecting combination of individual employees and tools). Even if the existence of the influence of the selection of production resources on the course of operations is signaled in the available works, the research so far has not focused on the measurable effect of such a solution that takes into account this phenomenon in scheduling. The proposed production scheduling methodology, including the influence of employees and tools, turned out to be more effective in terms of minimizing the maximum completion time and the cost of the production process compared to existing solutions. The efficiency of the new proposed scheduling methodology was assessed using examples of four technological processes. The research was carried out on the basis of a dedicated adaptation of the Monte Carlo optimization algorithm in order to determine the actual effect of the new solution. The algorithm itself is not an integral part of the proposed solution, and the universal methodology developed will ensure significant profit for any optimization algorithm correctly implemented.

The use of a zonal structure for energy markets across the globe is expanding; however the debate on how to effectively partition the grid into bidding zones is still open for discussion. One of the factors that needs to be addressed in the process of bidding zones’ delimitation is the transmission system operators control areas. Merging parts of different control areas into one bidding zone can lead to multiple problems, ranging from political, through grid security concerns, to reserve control issues. To address it, this paper presents a novel grid partitioning method aimed at bidding zones delimitation that is based on clustering the power grid using an extended version of the standard agglomerative clustering. The proposed solution adds additional clustering rules when constructing the dendrogram in order to take into account the control areas. The algorithm is applied to the data which represents the locational marginal prices obtained from optimal power flow analysis.

Forecasting promotion efficiency is an important issue in the fast-moving consumer goods sector. The objective of this paper is an analysis of the forecasting performance of two key performance indicators (KPI) used for the assessment of the sales process using machine learning methods. The authors present results of the experiments which were performed for 17 different products on real-life data from a large grocery company. In the paper feature extraction and construction methods are discussed also five different prediction algorithms are compared as well as the feature importance analyses are also provided. Out of the compared algorithms random forest leads and the feature importance are strongly related with the KPI.

We present a complete, fully automatic solution based on genetic algorithms for the optimization of discrete product placement and of order picking routes in a warehouse. The solution takes as input the warehouse structure and the list of orders and returns the optimized product placement, which minimizes the sum of the order picking times. The order picking routes are optimized mostly by genetic algorithms with multi-parent crossover operator, but for some cases also permutations and local search methods can be used. The product placement is optimized by another genetic algorithm, where the sum of the lengths of the optimized order picking routes is used as the cost of the given product placement. We present several ideas, which improve and accelerate the optimization, as the proper number of parents in crossover, the caching procedure, multiple restart and order grouping. In the presented experiments, in comparison with the random product placement and random product picking order, the optimization of order picking routes allowed the decrease of the total order picking times to 54%, optimization of product placement with the basic version of the method allowed to reduce that time to 26% and optimization of product placement with the methods with the improvements, as multiple restart and multi-parent crossover to 21%.

Instance selection and construction methods were originally designed to improve the performance of the k-nearest neighbors classifier by increasing its speed and improving the classification accuracy. These goals were achieved by eliminating redundant and noisy samples, thus reducing the size of the training set. In this paper, the performance of instance selection methods is investigated in terms of classification accuracy and reduction of training set size. The classification accuracy of the following classifiers is evaluated: decision trees, random forest, Naive Bayes, linear model, support vector machine and k-nearest neighbors. The obtained results indicate that for the most of the classifiers compressing the training set affects prediction performance and only a small group of instance selection methods can be recommended as a general purpose preprocessing step. These are learning vector quantization based algorithms, along with the Drop2 and Drop3 . Other methods are less efficient or provide low compression ratio.