Article
Computer Science, Information Systems
Meng Hu, Eric C. C. Tsang, Yanting Guo, Degang Chen, Weihua Xu
Summary: This paper proposes an attribute optimization algorithm based on overlap degree to accelerate attribute reduction and improve classification performance. By modeling the overlap degree of objects from different categories in advance, it can efficiently filter and optimize attributes for decision approximation.
INFORMATION SCIENCES
(2022)
Article
Computer Science, Artificial Intelligence
Daiwei Li, Haiqing Zhang, Tianrui Li, Abdelaziz Bouras, Xi Yu, Tao Wang
Summary: Two algorithms, JFCM-VQNNI and JFCM-FVQNNI, have been proposed in this research to achieve effective data imputation by considering clustering and uncertain information extraction when predicting missing values. Experimental results show that these two algorithms have higher imputation performance and reliability compared to traditional parameter-based imputation algorithms.
IEEE TRANSACTIONS ON FUZZY SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Qiong Chen, Mengxing Huang, Hao Wang, Guangquan Xu
Summary: Discretization is an important data preprocessing technique in data mining, especially in industrial control. However, traditional discretization methods have shortcomings, particularly in the preprocessing of high-resolution remote sensing big data, where necessary information is lost. This study proposes a discretization method for high-resolution remote sensing big data, which determines the membership degree of each pixel using linear decomposition and a fuzzy rough model, and selects discrete breakpoints using an adaptive genetic algorithm. The method achieves optimal discretization scheme in the shortest time by parallel computing the individual fitness of the population using a MapReduce framework.
IEEE TRANSACTIONS ON FUZZY SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Suyun Zhao, Zhigang Dai, Xizhao Wang, Peng Ni, Hengheng Luo, Hong Chen, Cuiping Li
Summary: This study introduces an accelerator for rule induction based on fuzzy rough theory, using consistency degree and key set to speed up the construction of rule classifiers. Experimental results show that the proposed accelerator performs significantly faster than unaccelerated rule-based classifier methods, especially on datasets with numerous instances.
IEEE TRANSACTIONS ON FUZZY SYSTEMS
(2021)
Article
Computer Science, Artificial Intelligence
Guanli Yue, Yanpeng Qu, Longzhi Yang, Changjing Shang, Ansheng Deng, Fei Chao, Qiang Shen
Summary: Fuzzy clustering is a method that uses partial memberships to decompose data into clusters, and it demonstrates comparable performance in knowledge exploitation when dealing with incomplete information. This article proposes a new fuzzy-rough intrigued harmonic discrepancy clustering (HDC) algorithm that effectively handles complex data distribution and improves clustering performance.
IEEE TRANSACTIONS ON FUZZY SYSTEMS
(2023)
Article
Computer Science, Information Systems
Wan-Lei Zhao, Hui Wang, Chong-Wah Ngo
Summary: This paper presents a simple yet effective solution for approximate k-nearest neighbor search and graph construction. The solution integrates graph construction and search tasks, and supports dynamic updates on the built graph.
IEEE TRANSACTIONS ON MULTIMEDIA
(2022)
Article
Computer Science, Information Systems
Dong He, Maureen Daum, Walter Cai, Magdalena Balazinska
Summary: DeepEverest is a system designed for efficient execution of interpretation by example queries over the activation values of a deep neural network. It includes an efficient indexing technique and a query execution algorithm with various optimizations, proven to be instance optimal. Experimental results demonstrate that DeepEverest significantly accelerates individual queries by up to 63 times using less storage than full materialization, consistently outperforming other methods on multi-query workloads simulating DNN interpretation processes.
PROCEEDINGS OF THE VLDB ENDOWMENT
(2021)
Article
Computer Science, Artificial Intelligence
Changzhong Wang, Yuhua Qian, Weiping Ding, Xiaodong Fan
Summary: This article proposes a novel criterion function for feature selection by redefining the concepts of fuzzy rough approximations using a class of irreflexive and symmetric fuzzy binary relations, and introducing the concept of inner product dependency to describe classification errors. Experimental results demonstrate the effectiveness of the proposed criterion function for datasets with a large overlap between different categories.
IEEE TRANSACTIONS ON FUZZY SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Chuan Luo, Sizhao Wang, Tianrui Li, Hongmei Chen, Jiancheng Lv, Zhang Yi
Summary: This paper presents parallel feature selection algorithms based on the rough hypercuboid approach to handle growing data volumes. Experimental results show that our algorithms are significantly faster than the original sequential counterpart while guaranteeing result quality. Moreover, the proposed algorithms can effectively utilize distributed-memory clusters to handle computationally demanding tasks.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2023)
Article
Computer Science, Artificial Intelligence
Meng Hu, Yanting Guo, Degang Chen, Eric C. C. Tsang, Qingshuo Zhang
Summary: The construction of fuzzy relations is crucial in fuzzy rough sets, and relations generated by soft distances are more robust. Two enhanced fuzzy similarity relations are proposed to improve attribute reduction and classification, using neighborhood and decision information. The algorithm is validated using gene expression profiles and demonstrates strong noise resistance and selection of tumor-related genes.
KNOWLEDGE-BASED SYSTEMS
(2023)
Article
Computer Science, Theory & Methods
Adnan Theerens, Chris Cornelis
Summary: Classical (fuzzy) rough sets are sensitive to noise, and we improve on this issue by introducing fuzzy quantifier-based fuzzy rough sets (FQFRS). We propose an intuitive fuzzy rough approximation operator that utilizes general unary and binary quantification models, and conduct a theoretical study of their properties. We also apply them to classification problems.
FUZZY SETS AND SYSTEMS
(2023)
Article
Transportation
Jinjun Tang, Xinshao Zhang, Weiqi Yin, Yajie Zou, Yinhai Wang
Summary: Accurate traffic flow analysis and modeling are crucial for ITS, and missing traffic data remains a critical issue. A hybrid method combining fuzzy rough set and fuzzy neural network is proposed for imputation of missing traffic data, showing superior performance compared to traditional methods across different time intervals and missing ratios.
JOURNAL OF INTELLIGENT TRANSPORTATION SYSTEMS
(2021)
Article
Computer Science, Artificial Intelligence
Paolo Avogadro, Matteo Alessandro Dominoni
Summary: This research presents an improved algorithm for exact discord search in time series analysis. By utilizing the warm-up process and similarity between closely related sequences, the algorithm (HST) reduces the search space significantly compared to HOT SAX. Numerical evidence suggests that the complexity of the discord search depends on the length of discords and the noise/signal ratio, and HST has proven to be more than 100 times faster than HOT SAX, making it a significant advancement in the field.
APPLIED INTELLIGENCE
(2022)
Article
Computer Science, Information Systems
Samet Memis
Summary: This paper redefines picture fuzzy soft matrices and proposes several distance measures. It also introduces a new kNN-based classifier and experimentally demonstrates its superiority over state-of-the-art classifiers. The findings show that the proposed method performs better in terms of accuracy and F1-score and is capable of modeling uncertainty and real-world problems.
Article
Computer Science, Artificial Intelligence
Saeed Zeraatkar, Fatemeh Afsari
Summary: Imbalanced data classification is a complex issue where traditional classifiers perform poorly due to imbalanced class distribution. This paper proposes a solution involving resampling the data in two phases, oversampling and undersampling, using robust extensions of KNN classifiers based on interval-valued fuzzy and intuitionistic fuzzy sets. The proposed method shows potential in handling both synthetic and real-world data sets with different levels of noise and borderlines.
EXPERT SYSTEMS WITH APPLICATIONS
(2021)
Article
Computer Science, Artificial Intelligence
Marko Palangetic, Chris Cornelis, Salvatore Greco, Roman Slowinski
Summary: This paper extends the fuzzy dominance-based rough set approach (DRSA) and explores the application of Ordered Weighted Average (OWA) operators. The theoretical properties of hybridizing OWA operators with fuzzy DRSA are examined, and the robustness of the standard fuzzy DRSA approach is experimentally compared with the OWA approach.
INTERNATIONAL JOURNAL OF APPROXIMATE REASONING
(2021)
Article
Computer Science, Theory & Methods
Marko Palangetic, Chris Cornelis, Salvatore Greco, Roman Slowinski
Summary: This paper discusses the importance of granular representations of crisp and fuzzy sets in rule induction algorithms based on rough set theory. It demonstrates that the OWA-based fuzzy rough set model, which has been successfully applied in various machine learning tasks, allows for a granular representation. The practical implications of this result for rule induction from fuzzy rough approximations are highlighted.
FUZZY SETS AND SYSTEMS
(2022)
Article
Computer Science, Artificial Intelligence
Oliver Urs Lenz, Daniel Peralta, Chris Cornelis
Summary: The study addresses the challenge of setting hyperparameters in one-class classification, introduces a new data descriptor ALP, evaluates it on a large collection of datasets, showing that ALP outperforms other descriptors, making it a good default choice.
PATTERN RECOGNITION
(2021)
Article
Computer Science, Artificial Intelligence
Francisco J. Baldan, Daniel Peralta, Yvan Saeys, Jose M. Benitez
Summary: Time series data is increasingly important in Big Data environments, with a lack of tools for time series processing identified as a challenge. A new approach based on time series features has shown significant progress in addressing time series problems, and has demonstrated outstanding performance on the latest datasets.
INTERNATIONAL JOURNAL OF COMPUTATIONAL INTELLIGENCE SYSTEMS
(2021)
Article
Computer Science, Artificial Intelligence
Oliver Urs Lenz, Daniel Peralta, Chris Cornelis
Summary: The study provides a thorough treatment of one-class classification with hyperparameter optimisation for five data descriptors. Experimental results show that the recent Malherbe-Powell proposal optimises the hyperparameters of all data descriptors most efficiently.
Article
Computer Science, Artificial Intelligence
Adnan Theerens, Oliver Urs Lenz, Chris Cornelis
Summary: Fuzzy rough set theory is a useful tool for dealing with inconsistent data in machine learning applications. The ordered weighted average (OWA) based fuzzy rough sets provide a solution to the problem of sensitivity to outliers in classical fuzzy rough sets. By extending it to Choquet-based fuzzy rough sets (CFRS), more flexibility and robustness can be achieved, including seamless integration of outlier detection algorithms.
INTERNATIONAL JOURNAL OF APPROXIMATE REASONING
(2022)
Article
Engineering, Electrical & Electronic
Timo De Waele, Adnan Shahid, Daniel Peralta, Anniek Eerdekens, Margot Deruyck, Frank A. M. Tuyttens, Eli De Poorter
Summary: To track the activities and performance of horses, inertial measurement units (IMUs) combined with machine learning algorithms are commonly used. A data-efficient algorithm is proposed that only requires 3 minutes of labeled calibration data. This algorithm achieved a 95% accuracy on datasets captured with leg-mounted IMUs and neck-mounted IMU. However, when the algorithm was calibrated on multiple horses and evaluated on unfamiliar horses, there was a 15% drop in classification accuracy.
IEEE SENSORS JOURNAL
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Daniel Peralta, Lin Tang, Maxim Lippeveld, Yvan Saeys
Summary: This paper studies the problem of overconfident predictions in fingerprint classification and proposes calibration methods and a modified search strategy to address it. Experimental results show that Dirichlet calibration can improve predicted class probabilities and reduce penetration rate while maintaining a good balance in performance.
2021 IEEE INTERNATIONAL CONFERENCE ON BIG DATA (BIG DATA)
(2021)
Proceedings Paper
Computer Science, Artificial Intelligence
Mauricio Restrepo, Chris Cornelis
Summary: The paper introduces functional dependency relations defined on the attribute set of an information system, and establishes basic relationships between functional dependency relations, attribute reduction, and closure operators. It demonstrates that reducts of an information system can be obtained from the maximal elements of a functional dependency relation using the partial order for dependencies.
ROUGH SETS (IJCRS 2021)
(2021)
Proceedings Paper
Computer Science, Artificial Intelligence
Olha Kaminska, Chris Cornelis, Veronique Hoste
Summary: Social media provide meaningful data for tasks like sentiment analysis and emotion recognition, which are often solved using deep learning methods. Due to the fuzzy nature of textual data, using classification methods based on fuzzy rough sets is considered. An approach for the SemEval-2018 emotion detection task using FRNN-OWA models and different text embedding methods achieved competitive results against more complex deep learning methods.
ROUGH SETS (IJCRS 2021)
(2021)
Proceedings Paper
Computer Science, Artificial Intelligence
Oliver Urs Lenz, Daniel Peralta, Chris Cornelis
Summary: The proposal suggests using interval-valued fuzzy sets to model concepts in datasets with missing values, expressing uncertainty through optimistic and pessimistic approximations. In a small experiment, it outperforms simple imputation methods like mean and mode on datasets with low missing value rates.
ROUGH SETS (IJCRS 2021)
(2021)