Article
Computer Science, Artificial Intelligence
Weihua Xu, Kehua Yuan, Wentao Li
Summary: The approximation space in rough set theory is crucial for handling uncertainties. The local rough set model, as an effective approach, improves learning efficiency by avoiding unnecessary information granule calculations. This paper investigates the dynamic approximation update mechanism for multigranulation data and proposes corresponding dynamic update algorithms based on the local generalized multigranulation rough set model.
APPLIED INTELLIGENCE
(2022)
Article
Computer Science, Information Systems
Ahmed Hamed, Ahmed Sobhy, Hamed Nassar
Summary: This approach tackles the challenges of processing a big incomplete information system by developing an efficient RST algorithm and distributing computational chores using the MapReduce framework. Experimental results validate the validity, accuracy, and efficiency of the approach, showing superior performance compared to similar approaches.
INFORMATION SCIENCES
(2021)
Article
Automation & Control Systems
Faryal Nosheen, Usman Qamar, Muhammad Summair Raza
Summary: This paper presents a parallel technique for calculating DRSA approximation sets. The proposed method directly computes approximations by following heuristic rules without calculating dominance positive or negative relations. The proposed parallel approach shows significant improvements in terms of execution time, memory consumption, and algorithmic complexity compared to traditional methods.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2022)
Article
Computer Science, Information Systems
Xin Yang, Miaomiao Li, Hamido Fujita, Dun Liu, Tianrui Li
Summary: In this paper, a method for improving the performance of attribute reduction in a dynamic data environment is proposed. By combining incremental technology and accelerated reduction strategy, the method utilizes stable attribute groups and matrix-based incremental mechanisms for reduction search and dynamic attribute reduction. Experimental results demonstrate the effectiveness of the proposed method in terms of stability, computational cost, and classification accuracy.
INFORMATION SCIENCES
(2022)
Article
Computer Science, Artificial Intelligence
Wentao Li, Weihua Xu, Xiaoyan Zhang, Jia Zhang
Summary: This paper discusses the application of a local rough set model based on dominance relation in ordered information systems, as well as the construction of multigranulation rough set models and the updating process of dynamic objects. Experimental evaluation demonstrates the superiority and effectiveness of the proposed dynamic updating approaches in ordered information systems.
ARTIFICIAL INTELLIGENCE REVIEW
(2022)
Article
Mathematics, Interdisciplinary Applications
Tareq M. Al-shami, Wen Qing Fu, E. A. Abo-Tabl
Summary: This paper presents rough approximations based on topology, using 8 types of E-neighborhoods to construct approximations of any subset X of U, and studying properties and relationships between these approximations. It also provides some easy-to-understand examples and compares our approximations with those in published literature.
Article
Computer Science, Artificial Intelligence
Muhammad Akram, Hafiza Saba Nawaz, Cengiz Kahraman
Summary: A rough set approximates a subset of a universal set based on a binary relation, and a Pythagorean fuzzy set provides information about truthness and falsity. This paper proposes a new combination of rough sets and Pythagorean fuzzy sets, called rough Pythagorean fuzzy sets, which can handle uncertainties in imprecise data. The manuscript presents a general framework for studying rough Pythagorean fuzzy approximations and discusses the properties of approximation operators induced from different binary relations. Algorithms for computing reduct family and rough Pythagorean fuzzy approximations are developed and applied to real-world examples.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Computer Science, Artificial Intelligence
Haoxiang Zhou, Wentao Li, Chao Zhang, Tao Zhan
Summary: In the new era of information society, studying incremental methods of calculating approximations and constructing incremental algorithms is wise to save computing time in rough methods. This study focuses on maintaining approximations dynamically in interval-valued ordered decision systems when the feature set and sample set increase or decrease. Incremental updating rules for four circumstances are obtained based on the matrix expression of approximations and dominated sets, and incremental algorithms are derived accordingly. Comparative experiments on calculation time verify the effectiveness and superiority of the proposed dynamic algorithms.
APPLIED INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Qian Zhou, Xiaojun Xie, Hua Dai, Weizhi Meng
Summary: Minimum vertex cover of hypergraphs is a variation of the widely studied minimum vertex covering problem. Existing algorithms for general graphs are not efficient enough for large-scale hypergraphs. To address this, we propose a novel rough set-based approach that combines rough set theory with stochastic local search algorithm. Experimental results show the advantages and limitations of our proposed approach.
NEURAL COMPUTING & APPLICATIONS
(2022)
Article
Computer Science, Artificial Intelligence
Jie Zhao, Daiyang Wu, JiaXin Wu, Eric W. K. See-To, Faliang Huang
Summary: The study proposes a novel approach, DIGAC, to improve DRSA by replacing object-based calculation with granule-based calculation. It effectively reduces time complexity and avoids repeated calculations using Dual Information Granules (DIGs). Experimental results show that the approach outperforms existing algorithms in terms of efficiency and stability, especially for large-scale and high-dimensional datasets.
APPLIED SOFT COMPUTING
(2023)
Article
Computer Science, Artificial Intelligence
Sandip Majumder, Samarjit Kar, Mihir Kr Chakraborty
Summary: This paper introduces a new concept of upper/lower approximations using rough membership functions and studies their algebraic properties. A game theoretic method is proposed to determine the thresholds and a comparison is made between these new approximations and other threshold-based approximations in rough set theory. The comparison suggests a potential application of the new concepts.
APPLIED INTELLIGENCE
(2023)
Article
Multidisciplinary Sciences
Fu Zhang, Weimin Ma, Hongwei Ma
Summary: In this article, the parameter of decision maker's familiarity with the attributes of the alternatives is introduced for the first time in dynamic multi-attribute group decision making to avoid drawbacks from inappropriate grouping. The combination with fuzzy soft rough set theory and dynamic multi-attribute-grouping decision making lead to a new decision model called dynamic chaotic multiple-attribute group decision making. An algorithm is provided to solve this model under a weighted T-spherical fuzzy soft rough set, achieving symmetry between decision evaluation and fuzzy information, and establishing a balanced relationship between decision makers and attributes.
Article
Computer Science, Theory & Methods
Linlin Xie, Guoping Lin, Jinjin Li, Yidong Lin
Summary: Attribute reduction, an essential challenge in pattern recognition, data mining, and knowledge discovery, can be improved by using a new measure called local information entropy.
FUZZY SETS AND SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Lei Ma, Chuan Luo, Tianrui Li, Hongmei Chen, Dun Liu
Summary: With the accumulation of interesting data in various application fields, incremental datasets are becoming more common. However, selecting informative attributes from dynamically changing datasets poses challenges. Therefore, an incremental processing mechanism is desired to update the attribute reducts efficiently. In this paper, a novel dynamic graph-based fuzzy rough attribute reduction approach is proposed to handle the maintenance of fuzzy rough attribute reduction in dynamic data, which outperforms existing methods in terms of speed and quality preservation.
INTERNATIONAL JOURNAL OF MACHINE LEARNING AND CYBERNETICS
(2023)
Article
Computer Science, Artificial Intelligence
Hua Mao, Shengyu Wang, Chang Liu, Gang Wang
Summary: Attribute reduction is a critical aspect of rough set theory in data analysis. Existing methods mainly focus on theories, leading to complexities in searching for attribute reducts. This paper proposes a visual method that overcomes this limitation and demonstrates its effectiveness in a conventional information system.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Statistics & Probability
Hongmei Chen, Jibo Wu, B. M. Golam Kibria
Summary: Yang and Huang have introduced a generalized Mahalanobis loss function for regression coefficient estimation. Through comparisons of different estimators based on the average loss criterion, conditions for the superiority of one estimator over the others are obtained. Theoretical results are illustrated with two numerical examples.
COMMUNICATIONS IN STATISTICS-THEORY AND METHODS
(2023)
Article
Computer Science, Theory & Methods
Zhihong Wang, Hongmei Chen, Zhong Yuan, Tianrui Li
Summary: Dimensionality reduction is an important preprocessing method for data analysis. The proposed hybrid dimensionality reduction method combines fuzzy rough set and linear discriminant analysis to improve interpretability and retain original feature information.
FUZZY SETS AND SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Jihong Wan, Hongmei Chen, Tianrui Li, Binbin Sang, Zhong Yuan
Summary: This study proposes a method for feature selection in data with uncertainty, fuzziness, and noise. A robust fuzzy rough set model is constructed to enhance the robustness and antinoise ability. Uncertainty measures are defined to analyze the interactivity and redundancy of features. Experimental results demonstrate the significance of the proposed method.
IEEE TRANSACTIONS ON FUZZY SYSTEMS
(2023)
Article
Computer Science, Theory & Methods
Chuan Luo, Sizhao Wang, Tianrui Li, Hongmei Chen, Jiancheng Lv, Zhang Yi
Summary: This article introduces a Rough Hypercuboid based Distributed Online Feature Selection (RHDOFS) method to address the challenges of Volume and Velocity in Big Data. It proposes a novel integrated feature evaluation criterion by exploring class separability in the boundary region. An efficient online feature selection method is developed for streaming features, and a parallel optimization mechanism is employed to accelerate the implementation. The algorithm is implemented on Apache Spark and demonstrates superior performance in comparison to other online feature selection algorithms.
IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS
(2023)
Article
Engineering, Electrical & Electronic
Hongmei Chen, Jian Wang, Lanyu Wang, Long LI, Honghui Deng, Xu Meng, Yongsheng Yin
Summary: This paper proposes a fully digital modulation calibration technique for channel mismatch of TIADC at any frequency. It estimates and stores the mismatch errors by pre-inputting a test signal in TIADC, and extracts the stored values for compensation when the input signal is at a special frequency. By adjusting the operation order and the order of correlation and modulation, the complexity of the proposed calibration algorithm is greatly reduced. The hardware consumption of filters in the calibration algorithm is also greatly reduced by introducing a CSD coding technique based on Horner's rule and sub-expression sharing. The FPGA verification results show significant improvements in SFDR and SNDR after calibration.
IEICE TRANSACTIONS ON ELECTRONICS
(2023)
Article
Computer Science, Artificial Intelligence
Zhaoyuan Wang, Shun Chen, Shenggong Ji, Zheyi Pan, Chuishi Meng, Junbo Zhang, Tianrui Li, Yu Zheng
Summary: Since renting a house is a low-frequency behavior, there is very limited data available for housing rental suggestions. This study proposes to investigate the issue using e-commerce data, as it shares the same users and can provide insights into their consumption attitudes and behaviors related to renting houses. However, integrating e-commerce data with geographic and traffic data poses a challenge due to the lack of supervised information. To address this, a pairwise approach is proposed to effectively utilize labeled users and a novel deep network called HouseCritic is developed to fuse features and evaluate user satisfaction. Experimental results using real-world data from Beijing, China demonstrate the effectiveness of the proposed approach, which is currently being deployed in an e-commerce company.
KNOWLEDGE-BASED SYSTEMS
(2023)
Article
Computer Science, Artificial Intelligence
Lishan Wu, Jie Hu, Fei Teng, Tianrui Li, Shengdong Du
Summary: This paper proposes an enhanced sample building method (ESNCSE) to construct positive and negative samples for text semantic matching tasks. Positive sample pairs are generated by randomly inserting punctuation marks into the original text, aiming to add noise simply and efficiently. To expand the number of negative samples without increasing calculation cost, the momentum contrast based on the sentence embedding method with soft negative sample (SNCSE) is utilized. The experimental results show that the average Spearman correlation coefficient is 79.74% for BERT-base and 80.64% for BERT-large in text semantic similarity task.
INTERNATIONAL JOURNAL OF MACHINE LEARNING AND CYBERNETICS
(2023)
Article
Computer Science, Artificial Intelligence
Xinyao Liu, Shengdong Du, Tianrui Li, Fei Teng, Yan Yang
Summary: With the rise of big data, the problem of data quality becomes more critical, especially the issue of missing values. Current research focuses on using neural network models like self-organizing mappings or automatic encoders for imputation. However, these methods struggle to discover interrelated and common features simultaneously. To address this, we propose a feature-fusion-enhanced autoencoder model for missing value filling, incorporating a hidden layer with de-tracking and radial basis function neurons to enhance feature learning. We also introduce a dynamic clustering-based missing value filling strategy for improved performance. Extensive experiments on thirteen datasets validate the effectiveness of our model.
APPLIED INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Zhengji Li, Yuexin Wu, Jin Yang, Jiangchuan Chen, Tianrui Li
Summary: This study addresses three significant issues in review-based recommender systems by proposing a flexible dual-optimizer network, utilizing BERT for contextual information extraction, and incorporating a time-varying feature extraction scheme. Extensive tests on benchmark datasets demonstrate the substantial performance increase of the proposed TADO model compared to existing techniques.
APPLIED INTELLIGENCE
(2023)
Article
Computer Science, Information Systems
Minbo Ma, Peng Xie, Fei Teng, Bin Wang, Shenggong Ji, Junbo Zhang, Tianrui Li
Summary: This study proposes a novel Hierarchical Spatio-temporal Graph Neural Network (HiSTGNN) that accurately predicts meteorological variables and stations over multiple time steps. The method constructs a hierarchical graph using an adaptive graph learning module, effectively capturing hidden spatial dependencies and diverse long-term trends. Dynamic interactive learning is introduced to facilitate information exchange between different hierarchical graphs.
INFORMATION SCIENCES
(2023)
Article
Computer Science, Information Systems
Zhihong Wang, Hongmei Chen, Xiaoling Yang, Jihong Wan, Tianrui Li, Chuan Luo
Summary: Dimensionality reduction is an important step in many learning methods to achieve optimal performance using discriminative features. This study proposes a fuzzy rough dimensionality reduction method that combines feature selection and feature extraction, and compares its performance with other algorithms, showing higher classification performance.
INFORMATION SCIENCES
(2023)
Article
Computer Science, Artificial Intelligence
Jihong Wan, Hongmei Chen, Tianrui Li, Min Li, Xiaoling Yang
Summary: Feature selection is crucial in data mining and knowledge discovery. Existing methods often overlook high-order interactions among variables, resulting in the loss of important dependency information. To address this, a robust knowledge metric approach is proposed to perceive and excavate the hidden latent information.
PATTERN RECOGNITION
(2023)
Article
Computer Science, Artificial Intelligence
Zhixuan Deng, Tianrui Li, Keyu Liu, Pengfei Zhang, Dayong Deng
Summary: This article redefines three probabilities and their corresponding mathematical expectations from the perspective of granular computing and investigates their properties. It also proposes a framework of feature selection algorithms based on probabilities and mathematical expectations. The theoretical analysis and experimental results show that probabilities and mathematical expectations have better performance than information entropy as feature selection criteria.
INTERNATIONAL JOURNAL OF MACHINE LEARNING AND CYBERNETICS
(2023)
Article
Computer Science, Artificial Intelligence
Wei Chen, Hongjun Wang, Zhiguo Long, Tianrui Li
Summary: Co-clustering methods utilize the correlation between samples and attributes to explore the co-occurrence structure in data, playing a significant role in gene expression analysis, image segmentation, and document clustering. Existing bipartite graph partition-based co-clustering methods have high time complexity and the same number of row and column clusters. To address these problems, this paper proposes a novel fast flexible bipartite graph model (FBGPC) that directly constructs the bipartite graph using the original matrix and utilizes the inflation operation to partition the graph and learn the co-occurrence structure. Hierarchical clustering is then used to obtain the clustering results based on the relationship of the co-occurrence structure.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2023)
Article
Computer Science, Artificial Intelligence
Zhaoyuan Wang, Zheyi Pan, Shun Chen, Shenggong Ji, Xiuwen Yi, Junbo Zhang, Jingyuan Wang, Zhiguo Gong, Tianrui Li, Yu Zheng
Summary: In this paper, a fine-grained, safe, and energy-efficient strategy to improve the efficiency of metro systems by dynamically scheduling dwell time for trains is proposed. A novel deep neural network called AutoDwell is introduced to tackle the challenges, optimizing the long-term rewards of dwell time settings through a reinforcement learning framework and capturing spatio-temporal correlations and interactions between trains. Extensive experiments on real-world datasets demonstrate the superior performance of AutoDwell in shortening passengers' overall travel time.
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING
(2023)