Abstract:In this article, a fuzzy floating centroids method (FFCM) is proposed, which uses a fuzzy strategy and the concept of floating centroids to enhance the performance of the neural-network classifier. The decision boundaries in the traditional floating centroids neural-network (FCM) classifier are "hard." These hard boundaries force a point, such as noisy or boundary point, to be assigned to a class exclusively, thereby frequently resulting in misclassification and influencing the performance of optimization methods to train the neural network. A fuzzy strategy combined with floating centroids is introduced to produce "soft" boundaries to handle noisy and boundary points, which increases the chance of discovering the optimal neural network during optimization. In addition, the FFCM adopts a weighted target function to correct the preference to majority classes for imbalanced data. The performance of FFCM is compared with ten classification methods on 32 benchmark datasets by using indicators: average <span class="mjpage"><svg xmlns:xlink="http://www.w3.org/1999/xlink" width="1.741ex" height="2.176ex" style="vertical-align: -0.338ex;" viewBox="0 -791.3 749.5 936.9" role="img" focusable="false" xmlns="http://www.w3.org/2000/svg"><g stroke="currentColor" fill="currentColor" stroke-width="0" transform="matrix(1 0 0 -1 0 0)"> <use xlink:href="#MJMATHI-46" x="0" y="0"></use></g></svg></span> -measure (Avg.FM) and generalization accuracy. Also, the proposed FFCM is applied to nondestructively estimate the strength grade of cement specimens based on microstructural images. In the experimental results, FFCM achieves the optimal generalization accuracy and Avg.FM on 17 datasets and 21 datasets, respectively; FFCM balances precision and recall better than its competitors for the estimation of cement strength grade.<svg xmlns="http://www.w3.org/2000/svg" style="display: none;"><defs id="MathJax_SVG_glyphs"><path stroke-width="1" id="MJMATHI-46" d="M48 1Q31 1 31 11Q31 13 34 25Q38 41 42 43T65 46Q92 46 125 49Q139 52 144 61Q146 66 215 342T285 622Q285 629 281 629Q273 632 228 634H197Q191 640 191 642T193 659Q197 676 203 680H742Q749 676 749 669Q749 664 736 557T722 447Q720 440 702 440H690Q683 445 683 453Q683 454 686 477T689 530Q689 560 682 579T663 610T626 626T575 633T503 634H480Q398 633 393 631Q388 629 386 623Q385 622 352 492L320 363H375Q378 363 398 363T426 364T448 367T472 374T489 386Q502 398 511 419T524 457T529 475Q532 480 548 480H560Q567 475 567 470Q567 467 536 339T502 207Q500 200 482 200H470Q463 206 463 212Q463 215 468 234T473 274Q473 303 453 310T364 317H309L277 190Q245 66 245 60Q245 46 334 46H359Q365 40 365 39T363 19Q359 6 353 0H336Q295 2 185 2Q120 2 86 2T48 1Z"></path></defs></svg>

Adaptive Clustering Neural Network Based on Fuzzy Covariance Matrix

Fuzzy Naive Bayes Classifier Based on Fuzzy Clustering

New fuzzy Kohonen clustering network

Adaptive Approach to Fuzzy Clustering

Study Of Artificial Neural Network Model Based On Fuzzy Clustering

Adaptive Nonstationary Fuzzy Neural Network

A Self-organizing Fuzzy Neural Networks

A fuzzy neural networks with structure learning

Reinforced Interval Type-2 Fuzzy Clustering-based Neural Network Realized Through Attention-based Clustering Mechanism and Successive Learning

Hierarchical Federated Learning with Adaptive Clustering on Non-IID Data

Deep Fuzzy Clustering Network With Matrix Norm Regularization

Design of double fuzzy clustering-driven context neural networks

NOVEL HYBRID CLUSTERING ALGORITHM AND ITS APPLICATION TO FAULT DIAGNOSIS

Robust Fuzzy Neural Network With an Adaptive Inference Engine

Weighted Adaptively Ensemble Clustering Method Based on Fuzzy Co-Association Matrix

Nonstationary fuzzy neural network based on FCMnet clustering and a modified CG method with Armijo-type rule

A Compensatory Fuzzy Neural Network Modeling Method Based on Clustering and Cultural Algorithms

Application of Neural Network Based on Fuzzy Clustering in the Classification Algorithm of Data Mining

Neural Network Soft Sensor Modeling Method Based on Adaptive Fuzzy Clustering

Improvement of Neural-Network Classifiers Using Fuzzy Floating Centroids

Adaptive fuzzy C-means clustering integrated with local outlier factor