Applied Intelligence最新文献

The popular block diagonal representation subspace clustering approach shows high effectiveness in dividing a high-dimensional data space into the corresponding subspaces. However, existing subspace clustering algorithms have some weaknesses in achieving high clustering performance. This paper presents a multi-geometric block diagonal representation subspace clustering with low-rank kernel (MBDR-LRK) method that includes two major improvements. First, as visual data often exists on a Riemannian manifold not captured by Euclidean geometry, we harness the multi-order data complementarity to develop a multi-geometric block diagonal representation (MBDR) subspace clustering. Secondly, the proposed MBDR-LRK approach ensures the low-rankness in the mapped space, by adapting the kernel matrix to a pre-defined one rather than relying on a fixed kernel as in traditional methods. The paper also presents details on the monotonic decrease of the objective function and the boundedness and convergence of the affinity matrix, and the experimental results prove the convergence of the proposed method. Based on the MATLAB development environment, the proposed MBDR-LRK algorithm outperforms other related algorithms and obtained an accuracy of 88.70% on the ORL (40 classes), 89.39% on the Extended Yale B (38 classes), 50.22% on the AR (100 classes) and 75.47% on the COIL (50 classes) datasets.

Multimodal sentiment analysis has received widespread attention from the research community in recent years; it aims to use information from different modalities to predict sentiment polarity. However, the model architecture of most existing methods is fixed, and data can only flow along an established path, which leads to poor generalization of the model to different types of data. Furthermore, most methods explore only intra- or intermodal interactions and do not combine the two. In this paper, we propose the Smart Routing Attention Network (SmartRAN). SmartRAN can smartly select the data flow path on the basis of the smart routing attention module, effectively avoiding the disadvantages of poor adaptability and generalizability caused by a fixed model architecture. In addition, SmartRAN includes the learning process of both intra- and intermodal information, which can enhance the semantic consistency of comprehensive information and improve the learning ability of the model for complex relationships. Extensive experiments on two benchmark datasets, CMU-MOSI and CMU-MOSEI, prove that the proposed SmartRAN has superior performance to state-of-the-art models.

Trajectory mapping onto a road network is a complex yet important task. This is because, in the presence of circular sections, Y-shaped intersections, and sections with elevated overlaps with the ground, the conditions of road networks become complicated. Consequently, trajectory mapping becomes challenging owing to the complexities of road networks and the noise generated by high positioning errors. In this study, in response to the difficulty in handling redundant noisy trajectory data in complex road network environments, a complex road network map-matching method based on trajectory structure extraction is proposed. The features of the structure are extracted from the original trajectory data to reduce the effects of redundancy and noise on matching. An adaptive screening candidate method is proposed using driver behavior to estimate the road density and reduce the matching time by selecting effective candidates. A spatiotemporal analysis function is redefined using speed and distance features, and a directional analysis function is proposed for use in combination with directional features to improve the matching accuracy of complex road networks. An experimental evaluation based on real-ground trajectory data collected using in-vehicle sensing devices is conducted to verify the effectiveness of the algorithm. Moreover, extensive experiments are performed on challenging real datasets to evaluate the proposed method, and its accuracy and efficiency are compared with those of two state-of-the-art map-matching algorithms. The experimental results confirm the effectiveness of the proposed algorithm.

High-utility itemset mining (HUIM) is a technique for extracting valuable insights from data. When dealing with sensitive information, HUIM can raise privacy concerns. As a result, privacy-preserving utility mining (PPUM) has become an important research direction. PPUM involves transforming quantitative transactional databases into sanitized versions that protect sensitive data while retaining useful patterns. Researchers have previously employed stochastic optimization methods to conceal sensitive patterns in databases through the addition or deletion of transactions. However, these approaches alter the database structure. To address this issue, this paper introduces a novel approach for hiding data with stochastic optimization without changing the database structure. We design a flexible objective function to let users restrict the negative effects of PPUM according to their specific requirements. We also develop a general strategy for establishing constraint matrices. In addition, we present a stochastic algorithm that applies the ant lion optimizer along with a hybrid algorithm, which combines both exact and stochastic optimization methods, to resolve the hiding problem. The results of extensive experiments are presented, demonstrating the efficiency and flexibility of the proposed algorithms.

In real-world complex group decision-making problems, preference inconsistency and opinion conflict are common and crucial challenges that need to be tackled. To promote consensus reaching, a novel group consensus reaching model is constructed considering individual satisfaction and group fairness. This study focuses on managing the group consensus reaching process based on flexible and adaptable information, modelled as distributed linguistic preference relations (DLPRs). First, a building process for DLPRs is discussed by integrating cumulative distribution functions converted from single linguistic term sets, hesitant fuzzy linguistic term sets, and comparative linguistic expressions. Furthermore, a two-stage consistency improvement method is proposed, which makes a trade-off between the frequency and magnitude of adjustments. Finally, we establish an improved group consensus model to balance individual satisfaction and group fairness, where individual satisfaction is measured by the deviation between the original and revised preferences and group fairness is measured by the deviation between individual satisfactions. The emergency response plan selection is conducted to show the validity and advantages of the proposed approach.

Graphical Abstract

Radiologists often inspect hundreds of two-dimensional computed-tomography (CT) images to accurately locate lesions and make diagnoses, by classifying and labeling the ribs. However, this task is repetitive and time consuming. To effectively address this problem, we propose a multi-axial rib segmentation and sequential labeling (MARSS) method. First, we slice the CT volume into sagittal, frontal, and transverse planes for segmentation. The segmentation masks generated for each plane are then reconstructed into a single 3D segmentation mask using binarization techniques. After separating the left and right rib volumes from the entire CT volume, we cluster the connected components identified as bones and sequentially assign labels to each rib. The segmentation and sequential labeling performance of this method outperformed existing methods by up to 4.2%. The proposed automatic rib sequential labeling method enhances the efficiency of radiologists. In addition, this method provides an extended opportunity for advancements not only in rib segmentation but also in bone-fracture detection and lesion-diagnosis research.

Physics-informed spatial-temporal discrete sequence learning networks have great potential in solving partial differential equations and time series prediction compared to traditional fully connected PINN algorithms, and can serve as the foundation for data-driven sequence prediction modeling and inverse problem analysis. However, such existing models are unable to deal with inverse problem scenarios in which the parameters of the physical process are time-varying and unknown, while usually failing to make predictions in continuous time. In this paper, we propose a continuous time series prediction algorithm constructed by the physics-informed graph neural ordinary differential equation (PGNODE). Proposed parameterized GNODE-GRU and physics-informed loss constraints are used to explicitly characterize and solve unknown time-varying hyperparameters. The GNODE solver integrates this physical parameter to predict the sequence value at any time. This paper uses epidemic prediction tasks as a case study, and experimental results demonstrate that the proposed algorithm can effectively improve the prediction accuracy of the spread of epidemics in the future continuous time.

Multimodal aspect-based sentiment analysis (MABSA) aims to predict the sentiment of aspect by the fusion of different modalities such as image, text and so on. However, the availability of high-quality multimodal data remains limited. Therefore, few-shot MABSA is a new challenge. Previous works are rarely able to cope with low-resource and few-shot scenarios. In order to address the above problems, we design a Few-shot Multimodal aspect-based sentiment analysis framework based on Contrastive Finetuning (FMCF). Initially, the image modality is transformed to the corresponding textual caption to achieve the entailed semantic information and a contrastive dataset is constructed based on similarity retrieval for finetuning in the following stage. Further, a sentence encoder is trained based on SBERT, which combines supervised contrastive learning and sentence-level multi-feature fusion to complete MABSA. The experiments demonstrate that our framework achieves excellent performance in the few-shot scenarios. Importantly, with only 256 training samples and limited computational resources, the proposed method outperforms fine-tuned models that use all available data on the Twitter dataset.

Cognitive and neurological impairments are very common, but only a small proportion of affected individuals are diagnosed and treated, partly because of the high costs associated with frequent screening. Detecting pre-illness stages and analyzing the progression of neurological disorders through effective and efficient intelligent systems can be beneficial for timely diagnosis and early intervention. We propose using Large Language Models to extract features from free dialogues to detect cognitive decline. These features comprise high-level reasoning content-independent features (such as comprehension, decreased awareness, increased distraction, and memory problems). Our solution comprises (i) preprocessing, (ii) feature engineering via Natural Language Processing techniques and prompt engineering, (iii) feature analysis and selection to optimize performance, and (iv) classification, supported by automatic explainability. We also explore how to improve Chatgpt’s direct cognitive impairment prediction capabilities using the best features in our models. Evaluation metrics obtained endorse the effectiveness of a mixed approach combining feature extraction with Chatgpt and a specialized Machine Learning model to detect cognitive decline within free-form conversational dialogues with older adults. Ultimately, our work may facilitate the development of an inexpensive, non-invasive, and rapid means of detecting and explaining cognitive decline.