Concept learning based on improved FCM- BiLSTM for fuzzy data classification and fusion

Abstract Concept-Cognitive Learning (CCL) is an effective concept learning approach that simulates human cognitive processes to facilitate knowledge discovery. However, existing CCL methods face two significant challenges. One is that existing models often assume accurate labels and ignore the possi...

Full description

Saved in:
Bibliographic Details
Main Authors: Jiaojiao Niu, Jiankun Zuo, Wenyan Tie
Format: Article
Language:English
Published: Nature Portfolio 2025-08-01
Series:Scientific Reports
Subjects:
Online Access:https://doi.org/10.1038/s41598-025-14821-3
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849763728584605696
author Jiaojiao Niu
Jiankun Zuo
Wenyan Tie
author_facet Jiaojiao Niu
Jiankun Zuo
Wenyan Tie
author_sort Jiaojiao Niu
collection DOAJ
description Abstract Concept-Cognitive Learning (CCL) is an effective concept learning approach that simulates human cognitive processes to facilitate knowledge discovery. However, existing CCL methods face two significant challenges. One is that existing models often assume accurate labels and ignore the possibility of inaccuracy, which may lead to decisions based on incorrect information. The other one is that the cognitive mechanisms in current CCL models do not account for the dependency relationships between objects, which limits the model’s ability to adapt to diverse datasets and handle complex relational patterns. Inspired by both fuzzy clustering and deep learning, this paper proposes a novel Concept-Cognitive Learning Model (FCLSCL), which integrates an improved Fuzzy C-means (FCM) and Bidirectional Long Short-Term Memory Network (BiLSTM). Specifically, an improved FCM is designed under the framework of fuzzy concept clustering to learn pseudo-concepts for each category and obtain the membership degree of each object to every category. Then, the weighted fuzzy concepts are introduced to capture the uncertainty of datasets by taking both membership degree and pseudo-concepts into account. For achieving concept classification tasks, the intents of the weighted fuzzy concept and fuzzy concept are concatenated together as inputs for BiLSTM to achieve bidirectional concept learning. Finally, experimental results, compared with several popular machine learning models, CCL methods and LSTM, demonstrate the effectiveness of the proposed FCLSCL.
format Article
id doaj-art-a0043cb13d6b4040bb25edf4a16f9253
institution DOAJ
issn 2045-2322
language English
publishDate 2025-08-01
publisher Nature Portfolio
record_format Article
series Scientific Reports
spelling doaj-art-a0043cb13d6b4040bb25edf4a16f92532025-08-20T03:05:18ZengNature PortfolioScientific Reports2045-23222025-08-0115111710.1038/s41598-025-14821-3Concept learning based on improved FCM- BiLSTM for fuzzy data classification and fusionJiaojiao Niu0Jiankun Zuo1Wenyan Tie2School of Computer Science, Yangtze UniversitySchool of Computer Science, Yangtze UniversitySchool of Computer Science, Yangtze UniversityAbstract Concept-Cognitive Learning (CCL) is an effective concept learning approach that simulates human cognitive processes to facilitate knowledge discovery. However, existing CCL methods face two significant challenges. One is that existing models often assume accurate labels and ignore the possibility of inaccuracy, which may lead to decisions based on incorrect information. The other one is that the cognitive mechanisms in current CCL models do not account for the dependency relationships between objects, which limits the model’s ability to adapt to diverse datasets and handle complex relational patterns. Inspired by both fuzzy clustering and deep learning, this paper proposes a novel Concept-Cognitive Learning Model (FCLSCL), which integrates an improved Fuzzy C-means (FCM) and Bidirectional Long Short-Term Memory Network (BiLSTM). Specifically, an improved FCM is designed under the framework of fuzzy concept clustering to learn pseudo-concepts for each category and obtain the membership degree of each object to every category. Then, the weighted fuzzy concepts are introduced to capture the uncertainty of datasets by taking both membership degree and pseudo-concepts into account. For achieving concept classification tasks, the intents of the weighted fuzzy concept and fuzzy concept are concatenated together as inputs for BiLSTM to achieve bidirectional concept learning. Finally, experimental results, compared with several popular machine learning models, CCL methods and LSTM, demonstrate the effectiveness of the proposed FCLSCL.https://doi.org/10.1038/s41598-025-14821-3Concept-cognitive learningFuzzy clusteringBiLSTMConcept classification
spellingShingle Jiaojiao Niu
Jiankun Zuo
Wenyan Tie
Concept learning based on improved FCM- BiLSTM for fuzzy data classification and fusion
Scientific Reports
Concept-cognitive learning
Fuzzy clustering
BiLSTM
Concept classification
title Concept learning based on improved FCM- BiLSTM for fuzzy data classification and fusion
title_full Concept learning based on improved FCM- BiLSTM for fuzzy data classification and fusion
title_fullStr Concept learning based on improved FCM- BiLSTM for fuzzy data classification and fusion
title_full_unstemmed Concept learning based on improved FCM- BiLSTM for fuzzy data classification and fusion
title_short Concept learning based on improved FCM- BiLSTM for fuzzy data classification and fusion
title_sort concept learning based on improved fcm bilstm for fuzzy data classification and fusion
topic Concept-cognitive learning
Fuzzy clustering
BiLSTM
Concept classification
url https://doi.org/10.1038/s41598-025-14821-3
work_keys_str_mv AT jiaojiaoniu conceptlearningbasedonimprovedfcmbilstmforfuzzydataclassificationandfusion
AT jiankunzuo conceptlearningbasedonimprovedfcmbilstmforfuzzydataclassificationandfusion
AT wenyantie conceptlearningbasedonimprovedfcmbilstmforfuzzydataclassificationandfusion