@article{Kim2024,
  author = {Kim, Jimin and Obregon, Josue and Park, Hoonseok and Jung, Jae Yoon},
  doi = {10.1016/j.rser.2024.114479},
  issn = {18790690},
  journal = {Renewable and Sustainable Energy Reviews},
  keywords = {Day-ahead solar power generation forecasting,Machine learning,Solar photovoltaic power plants,Transformer networks},
  month = aug,
  publisher = {Elsevier Ltd},
  title = {Multi-step photovoltaic power forecasting using transformer and recurrent neural networks},
  volume = {200},
  year = {2024}
}

Affordable and clean energy is an important UN sustainable development goal. Solar energy is more difficult to control than fossil fuels, highlighting the need for accurate solar power forecasts. This study develops three variants of the transformer networks, called PVTransNet, for a multi-step day-ahead photovoltaic power forecasting. The transformer networks use historical solar power generation, weather observation, weather forecast and solar geometry data as input to effectively predict next-day hourly power generation. In particular, the third variant model combines long short-term memory (LSTM) to transformer networks to supplement weather forecasts from the weather station. The combined model, PVTransNet-EDR, outperformed individual LSTM and other transformer models in the experiments conducted on data from two photovoltaic power plants. The model performed 48.3 % better, in mean absolute error, than simple LSTM in the power forecasting. Accurate solar power forecasting model is expected to be utilized for efficient energy storage and microgrid management, effective energy supply policy, and optimal plant location selection.

@article{Obregon2024,
  author = {Obregon, Josue and Jung, Jae Yoon},
  doi = {10.1007/S10845-022-02057-1},
  issn = {15728145},
  issue = {2},
  journal = {Journal of Intelligent Manufacturing},
  keywords = {Decision rules,Die-casting,Ensemble learning,Fault detection,Formal concept analysis,Interpretable machine learning},
  month = feb,
  pages = {521-537},
  publisher = {Springer},
  title = {Rule-based visualization of faulty process conditions in the die-casting manufacturing},
  volume = {35},
  year = {2024}
}

Die-casting is a popular manufacturing process that produces precise metal parts with excellent dimensional accuracy and smooth cast surfaces. Recently die-casting process condition data can be acquired to be used as input for machine learning techniques for fault detection. The rapid development of complex and accurate machine learning algorithms, such as tree ensembles and deep learning, allows the accurate detection of faulty products. However, interpreting and explaining black-box models is crucial in the die-casting industry because the predictions provided by the machine learning solution can be adopted in practice only after understanding the internal decision mechanism of the model. To solve this problem, rule extraction methods generate simple rule-based predictive models from complex tree ensembles. Nevertheless, rulesets may contain numerous complex rules with redundant conditions, and the standard structure of rulesets does not clearly show the hierarchical relationships and frequent interactions among their elements. For this reason, in this study, a visualization tool based on formal concept analysis, called RuleLat (Rule Lattice), is proposed, which generates simple visual representations of rule-based classifiers. The generated models depict the hierarchical relationships of interactions among conditions, rules, and predicted classes in a modified concept lattice that is easy to analyze and understand. To demonstrate the applicability of the proposed method, a case study using real-world manufacturing data collected from a die-casting company in Korea is presented. RuleLat is adopted as a tool for interpretable machine learning, and the process conditions of three types of defects (porosity, material, and imprint) are analyzed and discussed.

@article{Obregon2026,
  author = {Obregon, Josue and Han, Yu Ri and Ho, Chang Won and Mouraliraman, Devanadane and Lee, Chang Woo and Jung, Jae Yoon},
  doi = {10.1016/j.est.2023.106680},
  issn = {2352152X},
  journal = {Journal of Energy Storage},
  keywords = {Charge capacity estimation,Convolutional autoencoder,Deep learning,Electrochemical impedance spectroscopy,Lithium-ion batteries,State of health},
  month = apr,
  publisher = {Elsevier Ltd},
  title = {Convolutional autoencoder-based SOH estimation of lithium-ion batteries using electrochemical impedance spectroscopy},
  volume = {60},
  year = {2023}
}

The advancement of consumer electronics and electric vehicles requires heavy use of energy sources, particularly in the form of rechargeable batteries. Although lithium-ion batteries (LiBs) enable the use of such technologies owing to their high energy and power densities, estimating the state-of-health (SOH) of such batteries remains a challenge because of the various environmental operational conditions that affect the charging and discharging cycles of LiBs. In this study, we explore an approach that uses a convolutional autoencoder (CAE) for overcomplete feature extraction from electrochemical impedance spectroscopy (EIS) data. Subsequently, the extracted latent data representation is fed into a deep neural network (DNN) for battery capacity retention and SOH estimation. The proposed end-to-end deep learning-based architecture is called CAE-DNN. To prove the effectiveness of the proposed architecture, we conducted a series of experiments using a public dataset involving EIS spectra collected from fully charged LiBs cycled at different temperatures. The experimental results were compared with those of existing state-of-the-art methods, and with other classic machine learning methods. The results demonstrate that the proposed architecture extracts useful features in an unsupervised manner and estimates the SOH of LiBs more accurately than other baseline estimation methods.

@article{Obregon2023,
  author = {Obregon, Josue and Jung, Jae Yoon},
  doi = {10.1016/J.INFFUS.2022.08.021},
  issn = {1566-2535},
  journal = {Information Fusion},
  keywords = {Ensemble learning,Ensemble simplification,Explainable artificial intelligence,Interpretable machine learning,Rule extraction,tree ensembles},
  month = jan,
  pages = {355-381},
  publisher = {Elsevier},
  title = {RuleCOSI+: Rule extraction for interpreting classification tree ensembles},
  volume = {89},
  year = {2023}
}

Despite the advent of novel neural network architectures, tree-based ensemble algorithms such as random forests and gradient boosting machines still prevail in many practical machine learning problems in manufacturing, financial, and medical domains. However, tree ensembles have the limitation that the internal decision mechanisms of complex models are difficult to understand. Therefore, we present a post-hoc interpretation approach for classification tree ensembles. The proposed method, RuleCOSI+, extracts simple rules from tree ensembles by greedily combining and simplifying their base trees. Compared with its previous version, RuleCOSI, this new version can be applied to both bagging (e.g., random forest, RF) and boosting ensembles (e.g., gradient boosting machines, GBM) and run much faster for ensembles with hundreds of trees. To assess the performance and applicability of the method, empirical experiments were conducted using two bagging algorithms and four gradient boosting algorithms over 33 datasets. RuleCOSI+ could generate the best classification rulesets in terms of F-measure together with RuleFit for RF and GBM models of the datasets among five ensemble simplification algorithms, but the rulesets of RuleCOSI+ had, on average, less than half the size of those of RuleFit. Moreover, RuleCOSI+ had the best antecedent uniqueness rate (“UNIQ”) among the five algorithms, and had also ranked high in the number of rules (“NRULES”) and the rule reduction rate (“REDU”). In addition, the proposed method could reduce generalization errors in the simplified rulesets to obtain, on average, slightly better classification errors than original models of two bagging and three gradient boosting algorithms except CATBoost.

@article{Obregon2025,
  author = {Obregon, Josue and Hong, Jihoon and Jung, Jae-Yoon},
  doi = {10.1016/J.JMSY.2021.07.001},
  issn = {0278-6125},
  journal = {Journal of Manufacturing Systems},
  keywords = {Decision rules,Ensemble learning,Interpretable machine learning,Manufacturing quality condition,Plastic injection moulding},
  month = jul,
  pages = {392-405},
  publisher = {Elsevier},
  title = {Rule-based explanations based on ensemble machine learning for detecting sink mark defects in the injection moulding process},
  volume = {60},
  year = {2021}
}

Manufacturing quality control (QC) in plastic injection moulding is of the upmost importance since almost one third of plastic products are manufactured via the injection moulding process. Moreover, smart manufacturing technologies are enabling the generation of huge amounts of data in production lines. This data can be used for predicting the quality of manufactured plastic products using machine learning methods, allowing companies to save costs and improve their production efficiency. However, high-performance machine learning models are usually too complicated to be understood by human intuition. Therefore, we have introduced a rule-based explanations (RBE) framework that combines several machine learning interpretation methods to help to understand the decision mechanisms of accurate and complex predictive models – specifically tree ensemble models. These generated rules can be used to visually and easily understand the main factors that affect the quality in the manufacturing process. To demonstrate the applicability of RBE, we present two experiments with real industrial data gathered from a plastic injection moulding machine in a Singapore model factory. The collected datasets contain condition data for several manufacturing processes as well as the QC results for sink mark defects in the production of small plastic products. The experiments revealed that it is possible to extract meaningful explanations in the form of simple decision rules that are enhanced with partial dependence plots and feature importance rankings for a better understanding of the underlying mechanisms and data relationships of accurate tree ensembles.

@article{Obregon2019,
  author = {Obregon, Josue and Kim, Aekyung and Jung, Jae-Yoon},
  doi = {10.1016/j.eswa.2019.02.012},
  issn = {09574174},
  journal = {Expert Systems with Applications},
  keywords = {Boosting,Decision trees,Ensemble learning,Imbalanced classification,Rule extraction},
  title = {RuleCOSI: Combination and simplification of production rules from boosted decision trees for imbalanced classification},
  volume = {126},
  year = {2019}
}

© 2019 Elsevier Ltd In the field of machine learning, the problem of imbalanced classification arises when the class percentage on the data is unevenly distributed. Different strategies using boosting ensemble algorithms have shown improved results over the imbalanced classification problem by combining weak learners to produce a single strong learner. In particular, decision trees are often used as base learners in ensemble learning for classification or regression. However, boosting ensemble algorithms sometimes generate a large number of decision trees that could grow too large to be understandable and interpretable. Additionally, the use of weights adds more complexity to the final result. For this reason, in this paper, we present RuleCOSI, a novel method for combining and simplifying the output of an ensemble of binary decision trees into a single set of production rules. The proposed method takes into account the weight of each decision tree and using a combination matrix generates a single set of simplified production rules with performance comparable to that of the original boosting ensemble. In order to measure the performance and prove the applicability of the proposed method, we carried out an empirical validation using three different boosting algorithms over several well-known machine learning datasets as well as real-life data collected from a manufacturing company. The results of the algorithm are acceptable in most of the experiments reducing the complexity of the boosting ensemble output while maintaining a similar performance.

@article{Obregon2021,
  author = {Obregon, Josue and Song, Minseok and Jung, Jae Yoon},
  doi = {10.1109/ACCESS.2019.2906081},
  issn = {21693536},
  journal = {IEEE Access},
  keywords = {Community detection,Information flow,Network modularity,Process mining,Social networking services},
  pages = {48024-48036},
  publisher = {Institute of Electrical and Electronics Engineers Inc.},
  title = {InfoFlow: Mining Information Flow Based on User Community in Social Networking Services},
  volume = {7},
  year = {2019}
}

Online social networking services (SNSs) have emerged rapidly and have become huge data sources for social network analysis. The spread of the content generated by users is crucial in SNS, but there is only a handful of research works on information diffusion and, more precisely, information diffusion flow. In this paper, we propose a novel method to discover information diffusion processes from SNS data. The method starts preprocessing the SNS data using a user-centric algorithm of community detection based on modularity maximization with the purpose of reducing the complexity of the noisy data. After that, the InfoFlow miner generates information diffusion flow models among the user communities discovered from the data. The algorithm is an extension of a traditional process discovery technique called the Flexible Heuristics miner, but the visualization ability of the generated process model is improved with a new measure called response weight, which effectively captures and represents the interactions among communities. An experiment with Facebook data was conducted, and information flow among user communities was visualized. Additionally, a quality assessment of the models was carried out to demonstrate the effectiveness of the method. The final constructed models allowed us to identify useful information such as how the information flows between communities and information disseminators and receptors within communities.

@article{Kim2015,
  author = {Kim, K. and Obregon, J. and Jung, J.-Y.},
  doi = {10.1587/transinf.E97.D.811},
  issn = {17451361},
  issue = {4},
  journal = {IEICE Transactions on Information and Systems},
  keywords = {Context analysis,Facebook,Information flow,Process mining,Text mining},
  title = {Analyzing information flow and context for Facebook fan pages},
  volume = {E97-D},
  year = {2014}
}

As the recent growth of online social network services such as Facebook and Twitter, people are able to easily share information with each other by writing posts or commenting for another’s posts. In this paper, we firstly suggest a method of discovering information flows of posts on Facebook and their underlying contexts by incorporating process mining and text mining techniques. Based on comments collected from Facebook, the experiment results illustrate how the proposed method can be applied to analyze information flows and contexts of posts on social network services. Copyright © 2014 The Institute of Electronics, Information and Communication Engineers.

@inproceedings{Kim2025,
  author = {Kim, Minsik and Han, Young-Suk and Obregon, Josue and Jung, Jae-Yoon},
  doi = {10.1007/978-3-031-74482-2_11},
  isbn = {978-3-031-74482-2},
  issn = {2195-4364},
  journal = {International Conference on Flexible Automation and Intelligent Manufacturing},
  pages = {91-97},
  publisher = {Springer, Cham},
  title = {Discovering Dispatching Rules in a Semiconductor Fab Using Interpretable Machine Learning},
  url = {https://link.springer.com/chapter/10.1007/978-3-031-74482-2_11},
  year = {2024}
}

Recent studies have been conducted in the application of machine learning (ML)-based dispatching methods. Unfortunately, the internal dispatching behavior of such ML-based models is difficult to interpret. Therefore, this study transforms the ML-based model to a...

@inproceedings{DeSmedt2017,
  author = {Smedt, J. De and Broucke, S.K.L.M. Vanden and Obregon, J. and Kim, A. and Jung, J.-Y. and Vanthienen, J.},
  doi = {10.1007/978-3-319-58457-7_15},
  isbn = {9783319584560},
  issn = {18651348},
  journal = {Business Process Management Workshops},
  keywords = {DMN,Decision management,Decision mining},
  title = {Decision mining in a broader context: An overview of the current landscape and future directions},
  volume = {281},
  year = {2017}
}

© Springer International Publishing AG 2017. The term Decision Mining has been put forward in literature to cover numerous applications in a diverse set of contexts. In the business process management community, it typically reflects the way processes and data required for decision purposes in those processes are blended into one model during discovery. However, the upcoming field of decision modeling and management requires the term to be repositioned in order to obtain a better understanding of the interplay of processes and decisions. In this paper, the different approaches that are currently available are delineated and a case is made for a new type of decision mining: one that separates the control flow and decision perspective in a less stringent form compared to existing approaches.

@inproceedings{Kim2014,
  author = {Kim, A. and Obregon, J. and Jung, J.-Y.},
  doi = {10.1007/978-3-319-06257-0},
  isbn = {9783319062563},
  issn = {18651348},
  journal = {Business Process Management Workshops},
  keywords = {Decision tree,Performer recommendation,Process mining},
  title = {Constructing decision trees from process logs for performer recommendation},
  volume = {171 171 LN},
  year = {2014}
}

This paper demonstrates that the discovery technique using historical event logs can be extended to predict business performance and recommend performers for running instances. For the prediction and recommendation, we adopt decision trees, which is a decision support tool in management science. Decision trees are commonly used to help identify the most likely alternative to reach a goal. To provide effective performer recommendation, we use several filters with previous performers and key tasks to the decision tree. These filters allow for a suitable recommendation according to the characteristics of the processes. The proposed approach is implemented on ProM framework and it is then evaluated through an experiment using reallife event logs, taken from a Dutch Financial Institute. The main contribution of this paper is to provide a real-time decision support tool by recommendation of the best performer for a target performance indicator during process execution based on historical data. © Springer International Publishing Switzerland 2014.

@inproceedings{Obregon2017,
  author = {Obregon, J. and Kim, A. and Jung, J.-Y.},
  doi = {10.1007/978-3-319-02922-1_6},
  issn = {18651348},
  journal = {Asia Pacific Business Process Management},
  keywords = {Decision making,Decision tree,Process mining tool,Recommendation},
  title = {DTminer: A tool for decision making based on historical process data},
  volume = {159},
  year = {2013}
}

© Springer International Publishing Switzerland 2013. Process mining is a discipline that uses techniques to extract knowledge from event logs recorded by information systems in most companies these days. Among main perspectives of process mining, organizational and time perspectives focus on information about resources stored on the event logs and timing and frequency of the events, respectively. In this paper we introduce a method that combines organizational and time perspectives of process mining with a decision support tool called decision trees. The method takes the information of historical process data by means of an event log, generates a decision tree, annotates the decision tree with processing times, and recommends the best performer for a given running instance of the process. We finally illustrate the method through several experiments using a developed plug-in for the process mining framework ProM, first using synthetic data and then using a real-life event log.

@book_section{Obregon2022,
  author = {Obregon, Josue and Jung, Jae Yoon},
  doi = {10.1016/B978-0-323-85648-5.00011-6},
  isbn = {9780323856485},
  journal = {Human-Centered Artificial Intelligence: Research and Applications},
  keywords = {AdaBoost,Bagging,Boosting,Ensemble learning,Explainable artificial intelligence,Explainable machine learning,Interpretable machine learning,Tree ensembles},
  month = jan,
  pages = {51-72},
  publisher = {Academic Press},
  title = {Explanation of ensemble models},
  year = {2022}
}

Ensemble learning is a type of machine learning, typically supervised learning, that combines the decisions of multiple individual models to improve the classification or regression accuracy. Since their introduction 2 decades ago, ensemble models have been widely used not only in academia but also in practical applications, particularly in data science competitions such as Kaggle because they excel with tabular and structured data. However, understanding the decision mechanisms of such large models is a challenge. In the explainable artificial intelligence (XAI) literature, several studies have attempted to make ensemble models more transparent. Although deep learning has recently attracted significant attention in XAI research, it is still important to introduce methods that help explain ensemble models. This chapter introduces the problem of explaining the ensemble models in detail. Two of the most popular ensemble approaches, bagging and boosting, are first introduced, and then the main factors that make ensemble models difficult to explain are discussed. Later, a taxonomy of ensemble interpretation methods and some representative techniques for each category is also presented.