Despite the growing popularity of machine learning (ML), such solutions are often incomprehensible to employees and difficult to control. Addressing this issue, we discuss some essential problems of explainable ML applications in the fast-moving consumer goods (FMCG) market. This research puts forward a new approach to effective supply management by utilizing rough sets (RST), distance-based clustering, and dimensionality reduction techniques. In the presented case study, we aim to reduce the work done by experts by applying a single delivery plan to many similar points of sale (PoS). We achieve this objective by clustering vending machines based on historical sales patterns. To verify the feasibility of such an approach, we performed a series of experiments related to demand prediction on two data representations with various clustering techniques. The conducted experiments confirmed that, without losing quality in terms of MAE and RMSE, we could operate on PoS in an aggregate manner, thus reducing the workload of preparing delivery plans.

Large-scale deployment of AI in oncology is constrained less by standalone algorithmic performance than by system-level safety, accountability, interoperability, and regulation-aware governance. Grounded in approximately one year of practical pre-deployment work within the OnkoBot project, this paper specifies a deployment- and governance-first reference model for integrated oncology AI platforms under the EU AI Act and the Medical Device Regulation (MDR). The paper introduces Architecture for Medical AI Collaboration (AMAC), an implementation-neutral, system-level envelope that enforces strict online/offline separation between clinical operation and model/knowledge learning and evolution, gate-controlled releases via a Clinical Governance Gateway (CGG) with explicit human-in-the-loop (HITL) escalation, and tamper-evident auditability across clinical, technical, and interoperability boundaries. AMAC is anchored by the Community of Collaborative Evolving Medical Assistants (CEMA), a supervised multi-agent computational core that performs coordinated clinical reasoning under bounded autonomy. Concrete deliverables include: (i) a reference architecture outline with explicit responsibilities and auditable control points; (ii) a phase-gated deployment pathway (Preparation → Prototype → Pilot → Integration → AMAC operation) with required evidence packs, decision gates, and rollback/suspension mechanisms; and (iii) enforceable socio-technical gate criteria, including Socio-Technical Readiness Levels (STRL), readiness metrics, and accountability mapping (RACI). The model is intentionally non-normative and does not encode clinical guidelines; it provides a minimal, auditable governance architecture designed to make large-scale clinical AI integration feasible, controllable, and regulation-compatible in complex oncology environments.

This paper explores a rough set-based approach for supporting insightful reasoning in Intelligent Systems (ISs). The novelty lies in the introduction of a new concept for approximate reasoning processes based on granular computations. Although many rough set theory extensions developed over time focus on reasoning about (partial) set inclusion, these approximation spaces sometimes fall short when dealing with crucial aspects of approximate reasoning within ISs. Specifically, these systems aim to construct high-quality approximations of compound decision granules that represent solutions. Here, we present the basis for insightful reasoning realized through approximate reasoning processes grounded in granular computations. By doing so, we provide a sufficiently rich basis for designing IS problem solvers. This basis allows ISs to restructure or adapt their reasoning based on the generated granular computations, ultimately leading to high- quality granular solutions.

We present an approach based on the Interactive Granular Computing (IGrC) model as the basis for developing foundations of Complex Intelligent Systems, i.e., Intelligent Systems dealing with complex phenomena (IS’s). The generalization of GrC to IGrC was proposed to support the design of IS’s treated in IGrC as examples of complex granules (c-granules) with control. To make such systems successful, it is necessary to enable such systems to have continuous interaction with the physical world. The control of c-granules aims to properly implement the physical semantics of specified transformations of c-granules in the physical world. This implementation is based on the discovery of relevant configurations of physical objects, which provides the basis for perceiving relevant data about these objects and their interactions through the control of c-granules. Additionally, to create high-quality models that serve as the basis for the behavior of IS’s, these configurations must be adaptively adjusted by control to allow for the perception of relevant data used to induce those models. Unlike information granules from GrC, the correct implementation of c-granule transformations cannot be restricted to the abstract space. An important property of the IS’s discussed here is that they cannot be separated from interactions with the physical world. Hence, they cannot be confined to an abstract space. In particular, the relevance of IGrC in searching for rough computational building blocks for cognition is discussed. These computational building blocks are modeled by complex granules (c-granules) and their networks. It is also proposed to use IGrC as the basis for developing IS’s grounded on cognitive computing.

The article presents the RIONIDA learning algorithm based on combination of two widely-used empirical approaches: rule induction and instance-based learning for imbalanced data classification. The algorithm is a substantial extension of the well-known RIONA algorithm developed for balanced data. RIONIDA is relatively fast and significantly outperforms the state-of-the-art algorithms analysed in the paper.

The main aim of the paper is to discuss the architecture for the future Intelligent Systems (IS's) and Decision Support Systems (DS's) dealing with complex phenomena such as supporting medical decisions (diagnosis and therapy) and to emphasize challenges in designing such systems. More precisely, the paper presents arguments for developing a specialized computing model based on the interactive granular computing paradigm which can help to design IS's and DS's more close to the prototypes of real life decision making. In this regard, the paper brings to the fore different experiences faced during designing other medical IS's or DS's.As a starting step, the paper considers the experience of developing the OvuFriend platform and outlines some possible extension of it in the framework of the proposed architecture on the basis of Interactive Granular Computing (IGrC) model. Specifically, our attempt is to analyze a scheme, which is being used in the platform of OvuFriend for determining health risks and possibilities of a woman to conceive a child, from the perspective of IGrC. The target of the paper is two fold. Firstly, to show how the underlying AI algorithm of this scheme can be related with the notion of computing in the context of IGrC. Secondly, to identify possible extensions of the existing scheme so that it becomes more dynamic, interactive, and close to personalized medicine.

This paper is an attempt to develop a notion of interactive information system, which contrary to the classical notion of information system allows real physical interactions to play a role in the process of gathering information about the objects or situations. The proposed notion of interactive information system is grounded on the basic building blocks of Interactive Granular Computing (IGrC), known as complex granules (c-granules). The main point of departure of IGrC from other existing mathematical tools for modeling complex real physical phenomenon is to include a possibility of introducing a process of learning and perceiving through interactions initiated in the real physical world. Thus, the model is not completely restricted in a pure mathematical manifold. It has both the abstract module and implementational module connected by a notion of control.

This study discusses some essential problems of explainable machine learning applications in the FMCG market. The solution combines several machine learning techniques, including clustering, dimensionality reduction, rough set reducts, and rule-based explanations. We propose a novel approach to improve human-computer interaction with the XAI prototype method by generating human-readable cluster descriptions, emphasizing each cluster’s most discernible characteristics. To evaluate our method, we refer to the challenging task of demand prediction. The results confirmed that we could achieve five times better work performance without losing quality.

Informational granules are crucial objects in Interactive Granular Computing (IGrC). They are making it possible to link abstract objects and physical (not mathematical) objects in processes related to perceiving situations in the real world. The discussed IGrC computing model based on so-called complex granules is not purely abstract. Informational granules are responsible for the realization of the interaction between complex granules and the environment. The control of c-granules is responsible for the transformation of the current configuration of informational granules into the new one. The role of informational granules in perceiving the real world using c-granules is discussed.

We present the current research on the Interactive Granular Computing (IGrC) model and its relationships with the human-computer interaction processes. The existing rough set approaches to approximation of concepts grounded on (partial) containment of sets are extended, for the purposes of Intelligent Systems (IS's) interacting with human experts and complex phenomena, to approximations based on compound reasoning aiming to generate the right decisions about perceived situations in the real world. This paper is a step toward developing the foundations of such IS's. The decisions of IS's are constructed along the reasoning performed by complex granules (c-granules) which are responsible for creating interfaces between informational layers and physical layers of IS's, often synchronized or learnt from the reasoning performed by humans. Depending on applications, the decisions may take different forms, e.g.: compound decisions represented by the collections of decisions made in a given period of time, specifications of compound structural physical objects satisfying the wanted properties, (parameterized) learning algorithms generating high quality classifiers from samples of objects, pipelines of computations preserving some given constraints, etc. Both the construction of compound decisions and reasoning are performed over information perceived by means of c-granules used by IS's as interfaces for interactions with the physical world. Such interactions of c-granules are realized by the control layer (control, in short) of these granules. The discussed approximation of complex concepts in the context of IGrC is of fundamental importance for developing foundations of IS's aiming to solve complex problems.

The article concerns the well-known RIONA algorithm. We focus on the explainability property of this algorithm. The theoretical results, formulated and proved in the paper, show the relationships of the RIONA classifiers to both instance- and rule-based classifiers. In particular, we show the equivalence (relative to the classification) of the RIONA algorithm with the rule-based algorithm generating all consistent and maximally general rules from the neighbourhood of the test case.

We discuss the three-way rough set based approach for approximation of decision granules in Intelligent Systems (IS's). The novelty of the approach is in presenting a new concept of approximation space which is based on advanced reasoning tools. Many generalisations of the rough set approaches developed over the years are mainly concentrated around reasoning concerning (partial) inclusion of sets. However, such approximation spaces are not satisfactory to deal with important aspects of approximate reasoning by IS's aiming to construct of the high quality approximations of compound decision granules. We demonstrate a number of examples supporting this claim. In particular, in solving the considered in the paper problems are involved complex algorithmic optimization processes directed by reasoning tools supporting searching for (semi-)optimal approximations of decision granules in huge spaces. This paper is a step toward developing tools for derivation of granules supporting IS's in perceiving situations to a degree satisfactory for making the right decisions.

We present two novel theorems that allow for estimating the weight parameter in (weighted) kNN while dealing with imbalanced data. More precisely, the theorems for G-mean and F1-score are presented. The theorems assume 'totally random' distribution i.e. lack of dependency between features and class value. These results can be used for setting the default weights of classes for kNN-type classifiers, e.g. for imbalanced learning problems. Moreover, these theorems taken together illustrate the fact that without a precise specification of the particular performance measure we are interested in, the 'best classifier' term can be ambiguous or even misleading.

This paper is an attempt to present some grounds that a change is needed in the way of viewing mathematical tools, such as fuzzy sets, rough sets, in the context of classifying and approximating concepts pertaining to the real physical complex phenomenon. The paper argues in favour of developing models going beyond the pure mathematical manifold. The main idea is not to develop a theory only based on gathered data, rather to incorporate the methods of perception and real physical interactions through which the data is obtained. In this regard, a primary proposal has been put forward to model fuzzy sets and rough sets in the framework of Interactive Granular Computing (IGrC).

The theory of rough sets was founded by Zdzisław Pawlak as a framework for data and knowledge exploration. His seminal paper titled "Rough Sets" was published in 1982, in International Journal of Computer and Information Sciences. One of the key aspects that lets us use rough sets in practical scenarios is the notion of information system, which comes from even earlier Professor Pawlak’s works. Information systems are the means for data and knowledge representation. They constitute the input to rough set mechanisms aimed at computing approximations of concepts and deriving compacted, interpretable decision models. In particular, the fundamental notion of the indiscernibility relation is defined on the basis of a given information system. Accordingly, we discuss to what extent information systems can serve as the basis for intelligent systems. We claim that in many cases it is not enough to treat a data set – represented as an information system – as a purely abstract object with no linkage to the data origins. Oppositely, we should give ourselves a technical possibility to construct information systems dynamically, taking into account interaction with physical environments where the data comes from. With this respect, we refer to the notions of interactive granular computing and we generally consider together the paradigms of rough sets, information systems, and information granulation.

This paper is a continuation of our earlier works in establishing the need for introducing Interactive Granular Computing (IGrC) in developing Intelligent Systems (IS’s) and/or Decision Support Systems (DSS’s) dealing with complex phenomena. Among several crucial points, this paper argues in favour of the necessity to provide tools for learning models of complex vague concepts based on the perceived situations, where perception about the situation itself should be relativized based on the particular spatio-temporal windows of the physical world and real physical interactions among objects lying in the scope of those windows. The main idea is to develop a computing model which can link the abstract theory with its physical semantics in a way where the information about the world is grounded in the physical process of obtaining it and learning that information requires a proper implementation of interactions among real physical objects. The basic objects in IGrC are known as the complex granules (c-granules, for short). They make it possible to link the abstract and physical worlds, and help to realize the paths of judgments starting from generating a plan for obtaining sensory measurement or perception about a particular fragment of the physical world, to translating the plan to real physical interactions and verifying the properties obtained thereby with available knowledge. The c-granules, which are extended by information layers, are called informational c-granules (ic-granules, for short), and they can create the basis for modeling a notion of control conducting the whole process of computation over the c-granules. In this process an important role is played by so called implementational ic-granules responsible for the real physical realisation of the formal specification available in the information layer. Moreover, the networks of c-granules with distributed control are introduced and their role in IS’s dealing with complex phenomena is discussed.

This short paper is an attempt to clarify the role of Interactive Granular Computing (IGrC) as a computation model which respects that a real cognition about a real physical complex phenomenon and making decisions based on that cannot be formalized only being in the language of mathematics. In this regard, the paper focuses on presenting a real life example of computation where in order to move forward, without stumbling over the obstacles, a blind person needs to explore and learn the surrounding environment through interactions with the environment. The paper simply describes different components and features of IGrC model in the light of the concerned example and explains how this computing model has the potential to handle the grounding problem by bridging a connection between the abstract mathematical modeling and the real physical semantics.

This is a biogram of Professor Helena Rasiowa (1917–1994) one of the leading representatives of logicians from Warsaw. She was not only the great scientist but also the great human being. Rasiowa influenced numerous researchers from all over the world, especially by her results in algebraic logic, as well as by her great contribution for the mathematical community in all respect. She is also co-founder of the Pawlak-Rasiowa School of Artificial Intelligence (AI) (Jankowski and Skowron, Andrzej Mostowski and Foundational Studies, pp 106–143, 2008).

The problem of understanding intelligence is treated, by some prominent researchers, as the greatest problem of this century. In this article we justify that a decision support systems to be intelligent there is a need for developing new reasoning tools which can take into account the significance of the processes of sensory measurement, experience and perception about the concerned situations; i.e., understanding the process of perceiving a situation is also required for making relevant decisions. We discuss how such reasoning, called adaptive judgment, can be performed over objects interacting in the physical world using Interactive Granular Computing Model (IGrC). The basic objects in IGrC are called the complex granules (c-granules, for short). A c-granule is designed to link the abstract and physical worlds and to realize the paths of judgments starting from sensory measurement, experience to perception. Some c-granules are extended by information layers, called informational c-granules (ic-granules, for short); they can create the basis for modeling a notion of control conducting the whole process of computation over the c-granules.