International Journal of Pattern Recognition and Artificial Intelligence

To obtain the fault features of the bearing, a method based on variational mode decomposition (VMD), singular value decomposition (SVD) is proposed for fault diagnosis by Gath–Geva (G–G) fuzzy clustering. Firstly, the original signals are decomposed into mode components by VMD accurately and adaptively, and the spatial condition matrix (SCM) can be obtained. The SCM utilized as the reconstruction matrix of SVD can inherit the time delay parameter and embedded dimension automatically, and then the first three singular values from the SCM are used as fault eigenvalues to decrease the feature dimension and improve the computational efficiency. G–G clustering, one of the unsupervised machine learning fuzzy clustering techniques, is employed to obtain the clustering centers and membership matrices under various bearing faults. Finally, Hamming approach degree between the test samples and the known cluster centers is calculated to realize the bearing fault identification. By comparing with EEMD and EMD based on a recursive decomposition algorithm, VMD adopts a novel completely nonrecursive method to avoid mode mixing and end effects. Furthermore, the IMF components calculated from VMD include large amounts of fault information. G–G clustering is not limited by the shapes, sizes and densities in comparison with other clustering methods. VMD and G–G clustering are more suitable for fault diagnosis of the bearing system, and the results of experiment and engineering analysis show that the proposed method can diagnose bearing faults accurately and effectively.

In this paper, we propose generalized hesitant fuzzy soft sets by integrating generalized fuzzy soft sets with hesitant fuzzy sets. Then we investigate several fundamental properties of generalized hesitant fuzzy soft sets. In particular, a system of all the generalized hesitant fuzzy soft sets with narrow strict union and generalized strict intersection operators turns out to be a lattice. Meanwhile, a system of all the generalized hesitant fuzzy soft sets with generalized strict union and narrow strict intersection operators is a lattice. Based on generalized hesitant fuzzy soft sets, we provide an effective approach to decision making. Finally, three examples are presented to illuminate the feasibility of the method.

Moldy cores in apples are not initially obvious from the outside of the fruit, so developing methods to detect moldy cores is an important area of research in the apple industry. The objective of this study was to improve the ability of near-infrared spectrometry to detect moldy cores in apples. Transmission spectra were recorded for 200 apple samples in the range of 200–1100nm, and 140 and 60 samples were randomly selected as training and test sets, respectively. Signal de-noising was performed by wavelet thresholding based on the results of orthogonal experiments. The best wavelengths for discriminating between healthy and diseased apples were selected by a successive projection algorithm (SPA). The extracted wavelengths were used as the input in a back propagation artificial neural network (BP-ANN). Through these experiments, this study compared the correct recognition rates using different ratios of training to test numbers in the model, and functions in the hidden and output layers of the BP-ANN. The proposed method achieved the highest accuracies of 95.00% and 95.71% for the test and training sets, respectively. This method could be used to develop a portable instrument for detecting moldy cores in apples.

Learning-to-rank (LTR) is a very hot topic of research for information retrieval (IR). LTR framework usually learns the ranking function using available training data that are very cost-effective, time-consuming and biased. When sufficient amount of training data is not available, semi-supervised learning is one of the machine learning paradigms that can be applied to get pseudo label from unlabeled data. Cluster and label is a basic approach for semi-supervised learning to identify the high-density region in data space which is mainly used to support the supervised learning. However, clustering with conventional method may lead to prediction performance which is worse than supervised learning algorithms for application of LTR. Thus, we propose rank preserving clustering (RPC) with PLocalSearch and get pseudo label for unlabeled data. We present semi-supervised learning that adopts clustering-based transductive method and combine it with nonmeasure specific listwise approach to learn the LTR model. Moreover, each cluster follows the multi-task learning to avoid optimization of multiple loss functions. It reduces the training complexity of adopted listwise approach from an exponential order to a polynomial order. Empirical analysis on the standard datasets (LETOR) shows that the proposed model gives better results as compared to other state-of-the-arts.

With the advent of the Web era, data has exploded, with tens of thousands of textual data being generated every day. Traditional text sentiment analysis methods are mainly based on lexicon and machine learning-based methods. These methods show certain limitations as the data size increases. Here, we propose a new knowledge map design method based on convolutional neural network. For the academic literature data crawled from HowNet and Baidu Academic Website with the theme of computer, the corresponding ontology database, the fusion application of multiple data sources, and the mapping of different ontology libraries through data fusion are constructed for data sources in different fields. The global ontology library then uses the entity alignment and entity link methods for knowledge acquisition and fusion. Finally, the convolutional neural network is used for training and testing. The experimental results show that the subject search task can not only obtain the book through the convolution network, the effective academic literature in the question bank can also be used to obtain the relevance of the keyword in the search results, which verifies the effectiveness of the method.

In machine learning, there are many high-performance classifiers. However, because of lack of transparency, they are not able to explain the data in a human-friendly form. In this paper, Cumulative Fuzzy Class Membership Criterion (CFCMC), a recently proposed fuzzy modeling classifier, is modified and utilized for a novel approach of information extraction from the labeled data. This approach is able to explain the classifiability of the data in the form of semantics. Extracted semantics give information about the structure of the data and the similarities between classes. To get a relevant image of its classification performance, it is compared to three well-known and frequently used classifiers, which are considered as black boxes, namely, SVM, MLP, and kNN, and to a similar transparent approach, MF ARTMAP. To validate extracted semantics, they are compared to visualization of classified data and to confusion matrices generated during the evaluation of the created CFCMC models. The experimental result shows that CFCMC is not necessarily the best classifier, although, in most cases, it is not too far from the best performing methods. However, the semantical explanation potentially allows the classifier to be applied as a support for human decision processes in real-world problems.

As an art image, Thangka images have rich themes, various forms of expression, complex picture content and many layers of color representation. This paper mainly constructs a multi-core support vector machine (SVM) based on the information entropy feature-weighted radial basis kernel function. In this paper, the kernel function is optimized, and the feature reduction is performed by using the random forest feature selection algorithm with average accuracy degradation. Finally, the effective classification of the icon image and the mandala image in Thangka is realized. The research results provide support for the follow-up study of Thangka image annotation and retrieval.

The digital PCR instrument is a digital instrument for amplifying specific DNA fragments. The problem studied in this paper is the autofocus problem of its electronic imaging device. Based on the analysis of existing SOM neural network autofocus scheme, we propose an improved scheme-BP neural network for autofocus. It directly takes the SOM input and the actual focus position as the input and output of the BP neural network, which eliminates the process of prior classification and then corresponding to the focus matrix in the original SOM scheme, saving time. The experimental results show that the traditional autofocus method has good focusing effect, but the speed is slow, and the universality of the BP neural network autofocus scheme is not good enough, but within a good accuracy range, the speed is faster. Compared to traditional focusing methods, the autofocus scheme designed in this paper successfully achieves faster focusing speed for biochips.

Target detection is an important research topic in hyperspectral image processing, and has been the focus of research of many scholars. Compared with many target detection methods, the subspace target detection method is superior. This paper proposes a hyperspectral subspace target detection method LBSE-AMUSE based on the Local Background Subspace Estimation (LBSE) method and the AMUSE algorithm (Algorithm for Multiple Signals Extraction). HyMap airborne hyperspectral remote sensing data is used as the data source, and the performance between the LBSE and LBSE-AMUSE methods is compared by using quantitative indicators such as false alarm rate, detection rate and ROC curve. The experimental results show that the LBSE-AMUSE method has a simple structure, short running time and low false alarm rate compared with the LBSE method, which can suppress the background highlighting target to some extent.

The kernel function has been successfully utilized in the extreme learning machine (ELM) that provides a stabilized and generalized performance and greatly reduces the computational complexity. However, the selection and optimization of the parameters constituting the most common kernel functions are tedious and time-consuming. In this study, a set of new Hermit kernel functions derived from the generalized Hermit polynomials has been proposed. The significant contributions of the proposed kernel include only one parameter selected from a small set of natural numbers; thus, the parameter optimization is greatly facilitated and excessive structural information of the sample data is retained. Consequently, the new kernel functions can be used as optimal alternatives to other common kernel functions for ELM at a rapid learning speed. The experimental results showed that the proposed kernel ELM method tends to have similar or better robustness and generalized performance at a faster learning speed than the other common kernel ELM and support vector machine methods. Consequently, when applied to human action recognition by depth video sequence, the method also achieves excellent performance, demonstrating its time-based advantage on the video image data.

Face recognition helps in authentication of the user using remotely acquired facial information. The dynamic nature of face images like pose, illumination, expression, occlusion, aging, etc. degrades the performance of the face recognition system. In this paper, a new face recognition system using facial images with illumination variation, pose variation and partial occlusion is presented. The facial image is described as a collection of three complete connected graphs and these graphs are represented as symbolic objects. The structural characteristics, i.e. graph spectral properties, energy of graph, are extracted and embedded in a symbolic object. The texture features from the cheeks portions are extracted using center symmetric local binary pattern (CS-LBP) descriptor. The global features of the face image, i.e. length and width, are also extracted. Further symbolic data structure is constructed using the above features, namely, the graph spectral properties, energy of graph, global features and texture features. User authentication is performed using a new symbolic similarity metric. The performance is investigated by conducting the experiments with AR face database and VTU-BEC-DB multimodal database. The experimental results demonstrate an identification rate of 95.97% and 97.20% for the two databases.

Collecting data like location information is an essential part of concepts like the “IoT” or “Industry 4.0”. In the case of the development of a precise localization system and an integrated navigation system, indoor location technology receives more and more attention and has become a hot research topic. Common indoor location techniques are mainly based on wireless local area network, radio frequency tag, ZigBee technology, Bluetooth technology, infrared technology and ultra-wideband (UWB). However, these techniques are vulnerable to various noise signals and indoor environments, and also the positioning accuracy is easily affected by the complicated indoor environment. We studied the problem of real-time location tracking based on UWB in an indoor environment in this paper. We have proposed a combinational filtering algorithm and an improved Two-Way Ranging (ITWR) method for indoor real-time location tracking. The simulation results prove that the real-time performance and high accuracy of the presented algorithm can improve location accuracy. The experiment shows that the combinational algorithm and ITWR method which are applied to the positioning and navigation of the smart supermarket, have achieved quiet good results in positioning accuracy. The average positioning error is less than 10cm, some of the improvements can elevate the positioning accuracy by 17.5%. UWB is a suitable method for indoor real-time location tracking and has important theoretic value and practical significance.

Machine-learning-based software defect prediction (SDP) methods are receiving great attention from the researchers of intelligent software engineering. Most existing SDP methods are performed under a within-project setting. However, there usually is little to no within-project training data to learn an available supervised prediction model for a new SDP task. Therefore, cross-project defect prediction (CPDP), which uses labeled data of source projects to learn a defect predictor for a target project, was proposed as a practical SDP solution. In real CPDP tasks, the class imbalance problem is ubiquitous and has a great impact on performance of the CPDP models. Unlike previous studies that focus on subsampling and individual methods, this study investigated 15 imbalanced learning methods for CPDP tasks, especially for assessing the effectiveness of imbalanced ensemble learning (IEL) methods. We evaluated the 15 methods by extensive experiments on 31 open-source projects derived from five datasets. Through analyzing a total of 37504 results, we found that in most cases, the IEL method that combined under-sampling and bagging approaches will be more effective than the other investigated methods.

The digital processing of content resources has subverted the traditional paper content processing model and has also spread widely. The digital resources processed by text structure need to be structured and processed by professional knowledge, which can be saved as a professional digital content resource of knowledge base and provide basic metadata for intelligent knowledge service platform. The professional domain-based knowledge system construction system platform explored in this study is designed based on natural language processing. Natural language processing is an important branch of artificial intelligence, which is the application of artificial intelligence technology in linguistics. The system first extracts the professional thesaurus and domain ontology in the digital resources and then uses the new word discovery algorithm based on the label weight designed by artificial intelligence technology to intelligently extract and clean the new words of the basic thesaurus. At the same time, the relationship system between knowledge points and elements is established to realize the association extraction of targeted knowledge points, and finally the output content is enriched from knowledge points into related knowledge systems. In order to improve the scalability and universality of the system, the extended architecture of the thesaurus, algorithms, computational capabilities, tags, and exception thesaurus was taken into account when designing. At the same time, the implementation of “artificial intelligence $+$ manual assistance” was adopted. On the basis of improving the system availability, the experimental basis of the optimization algorithm is provided. The results of this research will bring an artificial intelligence innovation after the digitization to the publishing industry and will transform the content service into an intelligent service based on the knowledge system.

With the steady growth of the economy and the rapid development of modern industrial technology, the problem of environmental pollution has increased. To continue to develop, it is necessary to thoroughly implement the sustainable development strategy, and we must pay more attention to environmental issues. One of the important management tools implemented in China for environmental management is environmental quality monitoring and evaluation. Environmental quality monitoring can scientifically evaluate the environmental quality of a region, scientifically evaluate and forecast the environmental management and environmental engineering, and provide scientific basis for environmental management, environmental engineering, formulation of environmental standards, environmental planning, comprehensive prevention and control of environmental pollution, and ecological environment construction. This paper will discuss the basic principles of neural network and the implementation process of MATLAB and in the MATLAB software implementation and display process. At the same time, the results of different parameters are analyzed through experiments, and the network parameters are constantly adjusted to improve the accuracy of the evaluation results. Taking the regional environment as an example, two monitoring methods are proposed, and a variety of neural network models are used to analyze each prediction method. Case study results show that the latter method has a better prediction effect.