This study presents the initial results of the Myopia Risk Calculator (MRC) Consortium, introducing an innovative approach to predict myopia risk by using trustworthy machine-learning models. The dataset included approximately 7,945 records (eyes) from 3,989 children. We developed a myopia risk calculator and an accompanying web interface. Central to our research is the challenge of model trustworthiness, specifically evaluating the effectiveness and robustness of AI (Artificial Intelligence)/ML (Machine Learning)/NLP (Nat-ural Language Processing) models. We adopted a robust methodology combining Monte Carlo simulations with cross-validation techniques to assess model performance. Our experiments revealed that an ensemble of classifiers and regression models with Lasso regression techniques provided the best outcomes for predicting myopia risk. Future research aims to enhance model accuracy by integrating image and synthetic data, including advanced Monte Carlo simulations.

In this study, we extend our exploration of the AUC of Balance Accuracy Curve (BAC), a novel parameter we have developed that rivals the traditional metrics used for classification model evaluation, such as AUC of ROC and PR curves. BAC stands out for its straightforwardness, particularly in evaluating the overall efficacy of training systems amid varying degrees of class imbalance. Our current focus investigates the resilience of BAC against anomalies during the classification process, examining its behavior across multiple anomaly levels. For easier understanding, as a benchmark classifier we have employed the k-Nearest Neighbours (kNN) method, where we applied the most common distance metrics. The overall verification was conducted using real-world datasets selected from the UCI Repository, providing a practical context for our research.

In the evolving framework of granular computing, the continuous flow of data, often referred to as data streaming, presents both challenges and opportunities for the granulation process. Building upon the foundational works of Professor Zadeh and the granulation techniques based in rough set theory, this paper investigates the integration of data streaming with the concept-dependent granulation method. Recognizing the inherent challenges posed by the vastness and dynamism of streaming data, we explore the potential of adapting the concept-dependent granulation to accommodate and process these streams efficiently. Drawing inspiration from our previous works on random sampling and data decomposition, we introduce a novel approach that utilizes the real-time nature of data streams to enhance the granulation process. Our experimental studies, aim to evaluate the effectiveness of this method in terms of granulation quality and computational efficiency. Preliminary results suggest that integrating data streaming with concept-dependent granulation not only preserves the integrity of the granulated information but also offers significant advantages in processing large-scale dynamic data. In this work, we have verified the possibility of detecting the amount of data necessary to achieve the relevant classification efficiency, without having to process the entire data. This paper serves as a bridge, connecting the established methodologies of granulation with the emerging challenges of big data streaming, and sets the foundation for future research in this domain.

The granular computing approach as a paradigm in approximate reasoning deals with the processing of knowledge into granules that consist of entities similar in information content. Within the framework of rough set theory, proposed 40 years ago by Zdzislaw Pawlak and developed since then by many authors, granulation is an important area of research. Granulation techniques have found application in many areas of data mining in classification, feature selection, clustering and approximation processes for decision-making systems, among others. The current work is located in the development of approximation techniques for decision systems using rough inclusions proposed by Polkowski. Polkowski proposed the hypothesis that granules induced in a data set of a universe of objects should lead to new object representing them, and such granulated counterparts should preserve the information content of the data. This hypothesis has been verified in a number of research papers by Polkowski and Artiemjew. It has been proven that one of the best granulation techniques from the proposed family of methods is concept-dependent granulation. In which, on selected data, the degree of approximation of decision-making systems reaches more than 90 percent reduction in the size of training systems while maintaining the classification efficiency of the original training data. The undoubted disadvantage of this technique is the quadratic computational complexity - which forces us to look for a way to apply it on big data sets.The current work is one in a series of papers on the application of techniques used to deal with big data sets in the context of accelerating the concept-dependent granulation method. In this paper, we test whether granulation of training data divided into subgroups with post-granulation object fusion is competitive - in terms of running time and classification efficiency - to using a full training system in the granulation and classification process. The posed experimental problem is verified on a selected decision-making systems from UCI repository and using simple kNN classifier.

In this work, we propose a new parameter to study the effectiveness of classifiers -the AUC (area under curve) of the balanced accuracy curve (BAC) on data with different balance degrees -we compare its effectiveness with the popular AUC parameters for the ROC and PR curve.We use a global kNN classifier with typical metrics to verify the utility of the new parameter.BAC, ROC and PR curves generate similar results, the advantage of BAC is its simplicity of implementation and ease of interpretation of results.

In this work, we assess the classification capability of visualized multidimensional data used in the decision- making process. We want to investigate if classification carried out over a graphical representation of the tabular data allows for statistically greater efficiency than the dummy classifier method. To achieve this, we have used a convolutional neural network (ConvNet) as the base classifier. As an input into this model, we used data presented in the form of 2D curves resulting from the Parallel Coordinates Plot (PCP) visualization. Our initial results show that the topological arrangement of attributes, i.e., the shape formed by the PCP curves of individual data items, can serve as an effective classifier. Tests performed on three different real-world datasets from the UCI Machine Learning Repository confirmed that classification efficiency is significantly higher than in the case of dummy classification. The new method provides an interesting approach to the classificatio

One of the most popular families of techniques to boost classification are Ensemble methods. Random Forests, Bagging and Boosting are the most popular and widely used ones. This article presents a novel Ensemble Model, named Random Granular Reflections. The algorithm used in this new approach creates an ensemble of homogeneous granular decision systems. The first step of the learning process is to take the training system and cover it with random homogeneous granules (groups of objects from the same decision class that are as little indiscernible from each other as possible). Next, granular reflection is created, which is finally used in the classification process. Results obtained by our initial experiments show that this approach is promising and comparable with other tested methods. The main advantage of our new method is that it is not necessary to search for optimal parameters while looking for granular reflections in the subsequent iterations of our ensemble model.

This work extends the authors' contribution to the International Conference on Rough Sets and Current Trends in Computing (RSCTC 2008) held at the University of Akron in October 2008. It is dedicated to the topic of granular computing, formalized within the theory of rough mereology, as proposed by Polkowski; as an application of the idea of a granular reflection of data and of classifiers induced from it (Polkowski, 2005), we give an account of recent results in this area. A scheme for classifier construction based on factoring a classifier through a granular reflection of data; voting by granules of training objects; voting by granules of decision rules induced from the training set; voting by granules induced from a granular reflection of data, is presented here. In voting cases, voting is based on weights computed by means of rough inclusions induced from residual implications of continuous t-norms. The results show a high effectiveness of this approach as witnessed by the reported tests with some well--known data sets from University of California, at Irvine (UCI) repository whose results are compared against the standard rough set exhaustive classifier whose accuracy is indicated under the radius of “nil.” © 2011 Wiley Periodicals, Inc.