Search

In many practical applications data used for training a machine learning model and the deployment data does not always preserve the same distribution. Transfer learning and, in particular, domain adaptation allows to overcome this issue, by adapting the source model to a new target data distribution and therefore generalizing the knowledge from source to target domain. In this work, we present a method that makes the adaptation process more transparent by providing two complementary explanation mechanisms. The first mechanism explains how the source and target distributions are aligned in the latent space of the domain adaptation model. The second mechanism provides descriptive explanations on how the decision boundary changes in the adapted model with respect to the source model. Along with a description of a method, we also provide initial results obtained on publicly available, real-life dataset. © The Author(s) 2024., Funding: The paper is funded from the XPM project funded by the National Science Centre, Poland under the CHIST-ERA programme (NCN UMO2020/02/Y/ST6/00070) and the Swedish Research Council under grant CHIST-ERA19-XAI-012 and by a grant from the Priority Research Area (DigiWorld) under the Strategic Programme Excellence Initiative at Jagiellonian University.

In this work, we use an Evolutionary Algorithm (EA) to discover a novel Neural Network (NN) regression-based survival loss function with the aim of improving the C-index performance. Our contribution is threefold; firstly, we propose an evolutionary meta-learning algorithm SAGA$_{loss}$ for optimizing a neural-network regression-based loss function that maximizes the C-index; our algorithm consistently discovers specialized loss functions that outperform MSCE. Secondly, based on our analysis of the evolutionary search results, we highlight a non-intuitive insight that signifies the importance of the non-zero gradient for the censored cases part of the loss function, a property that is shown to be useful in improving concordance. Finally, based on this insight, we propose MSCE$_{Sp}$, a novel survival regression loss function that can be used off-the-shelf and generally performs better than the Mean Squared Error for censored cases. We performed extensive experiments on 19 benchmark datasets to validate our findings., As manuscript in thesis.

Recent machine learning studies on tabular data show that ensembles of decision tree models are more efficient and performant than deep learning models such as Tabular Transformer models. However, as we demonstrate, these studies are limited in scope and do not paint the full picture. In this work, we focus on how two dataset properties, namely dataset size and feature complexity, affect the empirical performance comparison between tree ensembles and Tabular Transformer models. Specifically, we employ a hypothesis-driven approach and identify situations where Tabular Transformer models are expected to outperform tree ensemble models. Through empirical evaluation, we demonstrate that given large enough datasets, deep learning models perform better than tree models. This gets more pronounced when complex feature interactions exist in the given task and dataset, suggesting that one must pay careful attention to dataset properties when selecting a model for tabular data in machine learning – especially in an industrial setting, where larger and larger datasets with less and less carefully engineered features are becoming routinely available. © The Author(s)

Mobility as a Service (MaaS) combines various modes of transportation to present mobility services to travellers based on their transport needs. This paper proposes a knowledge-based framework based on Artificial Intelligence (AI) to integrate various mobility data types and provide travellers with customized services. The proposed framework includes a knowledge acquisition process to extract and structure data from multiple sources of information (such as mobility experts and weather data). It also adds new information to a knowledge base and improves the quality of previously acquired knowledge. We discuss how AI can help discover knowledge from various data sources and recommend sustainable and personalized mobility services with explanations. The proposed knowledge-based AI framework is implemented using a synthetic dataset as a proof of concept. Combining different information sources to generate valuable knowledge is identified as one of the challenges in this study. Finally, explanations of the proposed decisions provide a criterion for evaluating and understanding the proposed knowledge-based AI framework. © 2023 by the authors.

Project within FFI: Accelerate the transition to sustainable road transport

Federated Learning (FL) is a setting where different clients collaboratively train a Machine Learning model in a privacy-preserving manner, i.e., without the requirement to share data. Given the importance of security and privacy in real-world applications, FL is gaining popularity in many areas, including predictive maintenance. For example, it allows independent companies to construct a model collaboratively. However, since different companies operate in different environments, their working conditions may differ, resulting in heterogeneity among their data distributions. This paper considers the fault identification problem and simulates different scenarios of data heterogeneity. Such a setting remains challenging for popular FL algorithms, and thus we demonstrate the considerations to be taken into account when designing federated predictive maintenance solutions.

The success of domain adaptation relies on high-quality labeled data from the source domain, which is a luxury setup for applied machine learning problems. This article investigates a particular challenge: the source labeled data are neither plentiful nor sufficiently representative. We studied the challenge of limited data with an industrial application, i.e., forklift truck activity recognition. The task is to develop data-driven methods to recognize forklift usage performed in different warehouses with a large scale of signals collected from the onboard sensors. The preliminary results show that using pseudo-labeled data from the source domain can significantly improve classification performance on the target domain in some tasks. As the real-world problems are much more complex than typical research settings, it is not clearly understood in what circumstance the improvement may occur. Therefore, we provided discussions regarding this phenomenon and shared several inspirations on the difficulty of understanding and debugging domain adaptation problems in practice. © 2023 IEEE., Funding: The Knowledge Foundation, Halmstad University, and Toyota Material Handling Europe

Time-To-Event (TTE) modeling using survival analysis in industrial settings faces the challenge of premature replacements of machine components, which leads to bias and errors in survival prediction. Typically, TTE survival data contains information about components and if they had failed or not up to a certain time. For failed components, the time is noted, and a failure is referred to as an event. A component that has not failed is denoted as censored. In industrial settings, in contrast to medical settings, there can be considerable uncertainty in an event; a component can be replaced before it fails to prevent operation stops or because maintenance staff believe that the component is faulty. This shows up as “no fault found” in warranty studies, where a significant proportion of replaced components may appear fault-free when tested or inspected after replacement. In this work, we propose an expectation-maximization-like method for discovering such premature replacements in survival data. The method is a two-phase iterative algorithm employing a genetic algorithm in the maximization phase to learn better event assignments on a validation set. The learned labels through iterations are accumulated and averaged to be used to initialize the following expectation phase. The assumption is that the more often the event is selected, the more likely it is to be an actual failure and not a “no fault found”. Experiments on synthesized and simulated data show that the proposed method can correctly detect a significant percentage of premature replacement cases., Som manuscript i avhandling/As manuscript in thesis.

The field of Explainable Predictive Maintenance (PM) is concerned with developing methods that can clarify how AI systems operate in the PM domain. One of the challenges of creating maintenance plans is integrating AI output with human decision-making processes and expertise. For AI to be helpful and trustworthy, fault predictions must be contextualized and easily comprehensible to humans. This involves providing tailored explanations to different actors depending on their roles and needs. For example, engineers can be connected to technical installation blueprints, while managers can evaluate system downtime costs, and lawyers can assess safety-threatening failures' potential liability. In many industries, black-box AI systems analyze sensor data to predict failures by detecting anomalies and deviations from typical behavior with impressive accuracy. However, PM is just one part of a broader context that aims to identify the most probable causes, develop a recovery plan, and estimate remaining useful life while providing alternative solutions. Achieving this requires complex interactions among various actors in industrial and decision-making processes. Our tutorial explores current trends, and promising research directions in Explainable AI (XAI) relevant to Explainable Predictive Maintenance (XPM), and future challenges and open issues on this topic. We will also present three case studies that highlight XPM's challenges in bus and train operations and steel factories. © 2023 Owner/Author., Funding: João Gama was financed by European Union’s Horizon Europe research and innovation programme under the Grant Agreement 101073874 (EMERITUS). Sławomir Nowaczyk and Sepideh Pashami were financed by Swedish Research Council under grant CHIST- ERA-19-XAI-012. Rita P. Ribeiro was financed by the CHIST-ERA grant CHIST-ERA- 19-XAI-012, and project CHIST-ERA/0004/2019 funded by FCT and also within project LA/P/0063/2020.

Accurate energy consumption prediction is crucial for optimizing the operation of electric commercial heavy-duty vehicles, e.g., route planning for charging. Moreover, understanding why certain predictions are cast is paramount for such a predictive model to gain user trust and be deployed in practice. Since commercial vehicles operate differently as transportation tasks, ambient, and drivers vary, a heterogeneous population is expected when building an AI system for forecasting energy consumption. The dependencies between the input features and the target values are expected to also differ across sub-populations. One well-known example of such a statistical phenomenon is the Simpson paradox. In this paper, we illustrate that such a setting poses a challenge for existing XAI methods that produce global feature statistics, e.g. LIME or SHAP, causing them to yield misleading results. We demonstrate a potential solution by training multiple regression models on subsets of data. It not only leads to superior regression performance but also more relevant and consistent LIME explanations. Given that the employed groupings correspond to relevant sub-populations, the associations between the input features and the target values are consistent within each cluster but different across clusters. Experiments on both synthetic and real-world datasets show that such splitting of a complex problem into simpler ones yields better regression performance and interpretability.

Machine Activity Recognition (MAR) can be used to monitor manufacturing processes and find bottlenecks and potential for improvement in production. Several interesting results on MAR techniques have been produced in the last decade, but mostly on construction equipment. Forklift trucks, which are ubiquitous and highly important industrial machines, have been missing from the MAR research. This paper presents a data-driven method for forklift activity recognition that uses Controller Area Network (CAN) signals and semi-supervised learning (SSL). The SSL enables the utilization of large quantities of unlabeled operation data to build better classifiers; after a two-step post-processing, the recognition results achieve balanced accuracy of 88% for driving activities and 95% for load-handling activities on a hold-out data set. In terms of the Matthews correlation coefficient for five activity classes, the final score is 0.82, which is equal to the recognition results of two non-domain experts who use videos of the activities. A particular success is that context can be used to capture the transport of small weight loads that are not detected by the forklift’s built-in weight sensor. © 2022 by the authors. Licensee MDPI, Basel, Switzerland.

One of the main challenges for predictive maintenance in real applications is the quality of the data, especially the labels. In this paper, we propose a methodology to filter out the misleading labels that harm the performance of Machine Learning models. Ideally, predictive maintenance would be based on the information of when a fault has occurred in a machine and what specific type of fault it was. Then, we could train machine learning models to identify the symptoms of such fault before it leads to a breakdown. However, in many industrial applications, this information is not available. Instead, we approximate it using a log of component replacements, usually coming from the sales or maintenance departments. The repair history provides reliable labels for fault prediction models only if the replaced component was indeed faulty, with symptoms captured by collected data, and it was going to lead to a breakdown. However, very often, at least for complex equipment, this assumption does not hold. Models trained using unreliable labels will then, necessarily, fail. We demonstrate that filtering misleading labels leads to improved results. Our central claim is that the same fault, happening several times, should have similar symptoms in the data; thus, we can train a model to predict them. On the contrary, replacements of the same component that do not exhibit similar symptoms will be confusing and harm the ML models. Therefore, we aim to filter the maintenance operations, keeping only those that can be used to predict each other. Suppose we can train a successful model using the data before a component replacement to predict another component replacement. In that case, those maintenance operations must be motivated by the same, or a very similar, type of fault. We test this approach on a real scenario using data from a fleet of sterilizers deployed in hospitals. The data includes sensor readings from the machines describing their operations and the service logs indicat

In classification problems, as the number of classes increases, correctly classifying a new instance into one of them is assumed to be more challenging than making the same decision in the presence of fewer classes. The essence of the problem is that using the learning algorithm on each decision boundary individually is better than using the same learning algorithm on several of them simultaneously. However, why and when it happens is still not well-understood today. This work’s main contribution is to introduce the concept of heterogeneity of decision boundaries as an explanation of this phenomenon. Based on the definition of heterogeneity of decision boundaries, we analyze and explain the differences in the performance of state of the art approaches to solve multi-class classification. We demonstrate that as the heterogeneity increases, the performances of all approaches, except one-vs-one, decrease. We show that by correctly encoding the knowledge of the heterogeneity of decision boundaries in a decomposition of the multi-class problem, we can obtain better results than state of the art decompositions. The benefits can be an increase in classification performance or a decrease in the time it takes to train and evaluate the models. We first provide intuitions and illustrate the effects of the heterogeneity of decision boundaries using synthetic datasets and a simplistic classifier. Then, we demonstrate how a real dataset exhibits these same principles, also under realistic learning algorithms. In this setting, we devise a method to quantify the heterogeneity of different decision boundaries, and use it to decompose the multi-class problem. The results show significant improvements over state-of-the-art decompositions that do not take the heterogeneity of decision boundaries into account. © 2013 IEEE.

Code Search is a practical tool that helps developers navigate growing source code repositories by connecting natural language queries with code snippets. Platforms such as StackOverflow resolve coding questions and answers; however, they cannot perform a semantic search through the code. Moreover, poorly documented code adds more complexity to search for code snippets in repositories. To tackle this challenge, this paper presents SIAMBERT, a BERT-based model that gets the question in natural language and returns relevant code snippets. The SIAMBERT architecture consists of two stages, where the first stage, inspired by Siamese Neural Network, returns the top K relevant code snippets to the input questions, and the second stage ranks the given snippets by the first stage. The experiments show that SIAMBERT outperforms non-BERT-based models having improvements that range from 12% to 39% on the Recall@1 metric and improves the inference time performance, making it 15x faster than standard BERT models., QC 20220930Part of proceedings: ISBN 978-166547126-8

In most industries, the working conditions of equipment vary significantly from one site to another, from one time of a year to another, and so on. This variation poses a severe challenge for data-driven fault identification methods: it introduces a change in the data distribution. This contradicts the underlying assumption of most machine learning methods, namely that training and test samples follow the same distribution. Domain Adaptation (DA) methods aim to address this problem by minimizing the distribution distance between training (source) and test (target) samples. However, in the area of predictive maintenance, this idea is complicated by the fact that different classes – fault categories – also vary across domains. Most of the state-of-the-art DA methods assume that the data in the target domain is complete, i.e., that we have access to examples from all the possible classes or faulty categories during adaptation. In reality, this is often very difficult to guarantee. Therefore, there is a need for a domain adaptation method that is able to align the source and target domains even in cases of having access to an incomplete set of test data. This paper presents our work in progress as we propose an approach for such a setting based on maintaining the geometry information of source samples during the adaptation. This way, the model can capture the relationships between different fault categories and preserve them in the constructed domain-invariant feature space, even in situations where some classes are entirely missing. This paper examines this idea using artificial data sets to demonstrate the effectiveness of geometry-preserving transformation. We have also started investigations on real-world predictive maintenance datasets, such as CWRU.

Mobility as a Service (MaaS) is a relatively new domain where new types of knowledge systems have recently emerged. It combines various modes of transportation and different kinds of data to present personalized services to travellers based on transport needs. A knowledge-based framework based on Artificial Intelligence (AI) is proposed in this paper to integrate, analyze, and process different types of mobility data. The framework includes a knowledge acquisition process to extract and structure data from various sources, including mobility experts and add new information to a knowledge base. The role of AI in this framework is to aid in automatically discovering knowledge from various data sets and recommend efficient and personalized mobility services with explanations. A scenario is also presented to demonstrate the interaction of the proposed framework’s modules.

This study applies a data-driven anomaly detection framework based on a Long Short-Term Memory (LSTM) autoencoder network for several subsystems of a public transport bus. The proposed framework efficiently detects abnormal data, significantly reducing the false alarm rate compared to available alternatives. Using historical repair records, we demonstrate how detection of abnormal sequences in the signals can be used for predicting equipment failures. The deviations from normal operation patterns are detected by analysing the data collected from several on-board sensors (e.g., wet tank air pressure, engine speed, engine load) installed on the bus. The performance of LSTM autoencoder (LSTM-AE) is compared against the multi-layer autoencoder (mlAE) network in the same anomaly detection framework. The experimental results show that the performance indicators of the LSTM-AE network, in terms of F1 Score, Recall, and Precision, are better than those of the mlAE network. © 2022 The Author(s), under exclusive license to Springer Nature Switzerland AG, Funding: The CHIST-ERA grant CHIST-ERA-19-XAI-012, project CHIST-ERA/0004/2019 funded by FCT - Fundação para a Ciência e Tecnologia and project 2020-00767 funded by Swedish Research Council., CHIST-ERA XPM

Survival Analysis models usually output functions (survival or hazard functions) rather than point predictions like regression and classification models. This makes the explanations of such models a challenging task, especially using the Shapley values. We propose SurvSHAP, a new model-agnostic algorithm to explain survival models that predict survival curves. The algorithm is based on discovering patterns in the predicted survival curves, the output of the survival model, that would identify significantly different survival behaviors, and utilizing a proxy model and SHAP method to explain these distinct survival behaviors. Experiments on synthetic and real datasets demonstrate that the SurvSHAP is able to capture the underlying factors of the survival patterns. Moreover, SurvSHAP results on the Cox Proportional Hazard model are compared with the weights of the model to show that we provide faithful overall explanations, with more fine-grained explanations of the sub-populations. We also illustrate the wrong model and explanations learned by a Cox model when applied to heterogeneous sub-populations. We show that a non-linear machine learning survival model with SurvSHAP can better model the data and provide better explanations than linear models., Funding: This research was funded by the CHIST-ERA grant CHIST-ERA-19-XAI-012 and CAISR+ project funded by the Swedish Knowledge Foundation.Som manuscript i avhandling/As manuscript in thesis.

Domain Adaptation (DA) aims to transfer knowledge from a source to a target domain by aligning their respective data distributions. In the unsupervised setting, however, this may cause the source and target samples of different classes to align to each other, consequently leading to negative transfer. Semi-Supervised Domain Adaptation (SSDA) tries to solve such class misalignment problem by exploiting a few sample labels in the target domain. This paper proposes a new SSDA method called Adversarial Contrastive Semi-Supervised Domain Adaptation (ACSSDA) which combines two objectives, optimized for the case where very few target sample labels are available, to learn a shared feature representation for both source and target domains. ACSSDA uses a domain classifier to ensure that the resulting feature space is domain agnostic. Simultaneously, Contrastive loss aims to pull together samples of the same class and push apart samples of different classes. This is shown to reduce class misalignment and negative transfer even with as little as a single labeled sample per class. We demonstrate the effectiveness of ACSSDA with experiments on several benchmark data sets. The results show the superiority of our method over state-of-the-art approaches.

Machine activity recognition is important for accurately esti- mating machine productivity and machine maintenance needs. In this paper, we present ongoing work on how to recognize activities of forklift trucks from on-board data streaming on the controller area network. We show that such recognition works across different sites. We first demon- strate the baseline classification performance of a Random Forest that uses 14 signals over 20 time steps, for a 280-dimensional input. Next, we show how a deep neural network can learn low-dimensional representa- tions that, with fine-tuning, achieve comparable accuracy. The proposed representation achieves machine activity recognition. Also, it visualizes the forklift operation over time and illustrates the relationships across different activities. © Springer Nature Switzerland AG 2020

This paper formulates the problem of predictive maintenance for complex systems as a hierarchical multi-class classification task. This formulation is useful for equipment with multiple sub-systems and components performing heterogeneous tasks. Often, the data available describes the whole system's operation and is not ideal for accurate condition monitoring. In this setup, specialized predictive models analyzing one component at a time rarely perform much better than random. However, using machine learning and hierarchical approaches, we can still exploit the data to build a fault isolation system that provides measurable benefits for technicians in the field. We propose a method for creating a taxonomy of components to train hierarchical classifiers that aim to identify the faulty component. The output of this model is a structured set of predictions with different probabilities for each component. In this setup, traditional machine learning metrics fail to capture the relationship between the performance of the models and its usefulness in the field.We introduce a new metric to evaluate our approach's benefits; it measures the number of tests a technician needs to perform before pinpointing the faulty component. Using a dataset from a real-case problem coming fro the automotive industry, we demonstrate how traditional machine learning performance metrics, like accuracy, fail to capture practical benefits. Our proposed hierarchical approach succeeds in exploiting the information in the data and outperforms non-hierarchical machine learning solutions. In addition, we can identify the weakest link of our fault isolation model, allowing us to improve it efficiently.

Nowadays, we can find machine learning (ML) applications in nearly every aspect of modern life, and we see that more developers are engaged in the field than ever. In order to facilitate the development of new ML applications, it would be beneficial to provide services that enable developers to share, access, and search for source code easily. A step towards making such a service is to cluster source code by functionality. In this work, we present COCLUBERT, a BERT-based model for source code embedding based on their functionality and clustering them accordingly. We build COCLUBERT using CuBERT, a variant of BERT pre-trained on source code, and present three ways to fine-tune it for the clustering task. In the experiments, we compare COCLUBERT with a baseline model, where we cluster source code using CuBERT embedding without fine-tuning. We show that COCLUBERT significantly outperforms the baseline model by increasing the Dunn Index metric by a factor of 141, the Silhouette Score metric by a factor of two, and the Adjusted Rand Index metric by a factor of 11., QC 20220530Part of proceedings ISBN 978-1-6654-4337-1

Batteries are a safety-critical and the most expensive component for electric vehicles (EVs). To ensure the reliability of the EVs in operation, it is crucial to monitor the state of health of those batteries. Monitoring their deterioration is also relevant to the sustainability of the transport solutions, through creating an efficient strategy for utilizing the remaining capacity of the battery and its second life. Electric buses, similar to other EVs, come in many different variants, including different configurations and operating conditions. Developing new degradation models for each existing combination of settings can become challenging from different perspectives such as unavailability of failure data for novel settings, heterogeneity in data, low amount of data available for less popular configurations, and lack of sufficient engineering knowledge. Therefore, being able to automatically transfer a machine learning model to new settings is crucial. More concretely, the aim of this work is to extract features that are invariant across different settings. In this study, we propose an evolutionary method, called genetic algorithm for domain invariant features (GADIF), that selects a set of features to be used for training machine learning models, in such a way as to maximize the invariance across different settings. A Genetic Algorithm, with each chromosome being a binary vector signaling selection of features, is equipped with a specific fitness function encompassing both the task performance and domain shift. We contrast the performance, in migrating to unseen domains, of our method against a number of classical feature selection methods without any transfer learning mechanism. Moreover, in the experimental result section, we analyze how different features are selected under different settings. The results show that using invariant features leads to a better generalization of the machine learning models to an unseen domain., Som manuscript i avhandling/As manuscript in thesis

Ensemble learning methods combine multiple models to improve performance by exploiting their diversity. The success of these approaches relies heavily on the dissimilarity of the base models forming the ensemble. This diversity can be achieved in many ways, with well-known examples including bagging and boosting. It is the diversity of the models within an ensemble that allows the ensemble to correct the errors made by its members, and consequently leads to higher classification or regression performance. A mistake made by a base model can only be rectified if other members behave differently on that particular instance, and provide the aggregator with enough information to make an informed decision. On the contrary, lack of diversity not only lowers model performance, but also wastes computational resources. Nevertheless, in the current state of the art ensemble approaches, there is no guarantee on the level of diversity achieved, and no mechanism ensuring that each member will learn a different decision boundary from the others. In this paper, we propose a parallel orthogonal deep learning architecture in which diversity is enforced by design, through imposing an orthogonality constraint. Multiple deep neural networks are created, parallel to each other. At each parallel layer, the outputs of different base models are subject to Gram–Schmidt orthogonalization. We demonstrate that this approach leads to a high level of diversity from two perspectives. First, the models make different errors on different parts of feature space, and second, they exhibit different levels of uncertainty in their decisions. Experimental results confirm the benefits of the proposed method, compared to standard deep learning models and well-known ensemble methods, in terms of diversity and, as a result, classification performance. © 2021 The Author(s). Published by Elsevier Ltd., CAISR+

The advancements of the telematics and connectivity solutions have provided new opportunities for the field of predictive maintenance. The number of sensors installed on a vehicle is increasing over time, and manufacturers are looking for new ways to improve the uptime of their fleet while at the same time reducing the costs related to unexpected breakdowns. The nature of the aggregated data from vehicles is sequential, and it is interesting to investigate existing methods for modeling partially observable state sequences to detect common patterns of failure. In this paper, we introduce a new approach for predicting turbocharger failures of Volvo trucks. The first step of the method deals with modeling a sequence of readouts from each vehicle using a Markov process. To do so, we identify the most informative signals and then employ spatial similarity clustering on the readouts. We interpret each cluster as a Markov state and further convert the history of a truck into a trajectory of states. This trajectory is then aligned with repairs information to form a standard sequence labeling problem. Finally, we train a hidden Markov model (HMM) classifier for assessing the health condition of the equipment. Empirical evaluations obtained on our realworld dataset of trucks suggest that the proposed method improves the AUC score of the final system up to 6% for predicting failures of a turbocharger.

The data shared over the Internet tends to originate from ubiquitous and autonomous sources such as mobile phones, fitness trackers, and IoT devices. Centralized and federated machine learning solutions represent the predominant way of providing smart services for users. However, moving data to central location for analysis causes not onlymany privacy concerns, but also communication overhead. Therefore, incertain situations machine learning models need to be trained in a collaborative and decentralized manner, similar to the way the data is originally generated without requiring any central authority for data or modelaggregation. This paper presents a decentralized and adaptive k-means algorithm that clusters data from multiple sources organized in peer-to-peer networks. Our algorithm allows peers to reach an approximation of the global model without sharing any raw data. Most importantly, we address the challenge of decentralized clustering with skewed non-IID data and asynchronous computations by integrating HyperLogLog counters with k-means algorithm. Furthermore, our clustering algorithm allows nodes to individually determine the number of clusters that fits their local data. Results using synthetic and real-world datasets show that our algorithm outperforms state-of-the-art decentralized k-means algorithms achieving accuracy gain that is up-to 36%. © Springer Nature Switzerland AG 2020., This research has been conducted within the BIDAF: A Big Data Analytics Framework for a Smart Society (http://bidaf.sics.se/).