Expert Systems with Applications最新文献

{"title":"Information-assisted and sentiment relation-driven for aspect-based sentiment analysis","authors":"Tiquan Gu ,&nbsp;Zhenzhen He ,&nbsp;Zhe Li ,&nbsp;Yaling Wan","doi":"10.1016/j.eswa.2025.127308","DOIUrl":"10.1016/j.eswa.2025.127308","url":null,"abstract":"<div><div>Aspect-based sentiment analysis aims to extract fine-grained sentiment information for different aspects in a review sentence. While existing methods have explored various ways to model the relationships between aspect terms and context, they often overlook the sentiment connection between aspect terms and the sentiment expressed by the review sentence itself. The sentiment tone of the sentence typically reflects the reviewer’s preferences and focus on certain aspects. Capturing this sentiment relationship allows the model to gain a more comprehensive understanding of the reviewer’s emotional experience and attitude. In this paper, we propose an information-assisted and sentiment relation-driven multitask learning network (IASD-ML) to address this gap. We define and label the relationship between aspect polarity and sentence polarity, treating it as an auxiliary task to learn the reviewer’s emotional context and the emotional associations between aspects and sentences. To the best of our knowledge, this is the first attempt to extract the sentiment relationship between aspect terms and sentence sentiment as an auxiliary classification task. Furthermore, relying solely on coarse-grained emotional cues from context is often insufficient to fully capture semantic and implicit relationships. To address this, we incorporate external commonsense path information to assist in extracting fine-grained sentiment cues and background information. Specifically, we use an external sentiment lexicon to label emotional words in the sentence, treating aspect terms as head entities and emotional words as tail entities, and retrieve commonsense path information from the ConceptNet knowledge base. By combining word dependencies with commonsense path information, we construct a commonsense aware graph network to further strengthen the emotional connections between aspect terms and sentiment words. Experimental results on benchmark datasets demonstrate that our approach has a solid competitive advantage.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"278 ","pages":"Article 127308"},"PeriodicalIF":7.5,"publicationDate":"2025-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143748558","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Quantification and mitigation of border-level localization deviation for object detectors","authors":"Chaojun Lin,&nbsp;Ying Shi,&nbsp;Changjun Xie,&nbsp;Mengqi Li","doi":"10.1016/j.eswa.2025.127435","DOIUrl":"10.1016/j.eswa.2025.127435","url":null,"abstract":"<div><div>Environmental perception is a critical module of automated driving systems for detecting obstacles and providing a decision-making basis for planning and control modules. In recent years, many localization deviation correction methods have emerged. However, these methods rarely study the border-level deviations, and current evaluation metrics cannot quantify the border-level deviation. To solve this problem, we model the maximum border localization deviation as a Gaussian distribution and propose a series of quantitative metrics to represent the border localization deviation as the absolute sum of the mean and variance of the distribution. Based on it, we proposed a predictive distribution fusion module to embed the predictive information into detection head networks, making the heads rethink and learn to reduce deviation. Experimental results demonstrate that our method can be flexibly integrated with various state-of-the-art detectors, further improving detection accuracy by approximately 1.0 mAP and enhancing the overall localization quality score by more than 6%. At an inference speed of 26.7 FPS, it achieves a detection accuracy of 43.3 mAP in urban road environments. The code and trained models are available at <span><span>https://github.com/unbelieboomboom/RefineHead</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"278 ","pages":"Article 127435"},"PeriodicalIF":7.5,"publicationDate":"2025-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143748565","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Large language model for interpreting research policy using adaptive two-stage retrieval augmented fine-tuning method","authors":"Runtao Ren ,&nbsp;Jian Ma ,&nbsp;Zhimin Zheng","doi":"10.1016/j.eswa.2025.127330","DOIUrl":"10.1016/j.eswa.2025.127330","url":null,"abstract":"<div><div>Accurate interpretation of scientific funding policies is crucial for government funding agencies and research institutions to make informed decisions and allocate research funds effectively. However, current large language model (LLM)-based systems often generate responses without references, leading to a lack of interpretability needed for policy enforcement. This study introduces the Adaptive Two-stage Retrieval Augmented Fine-Tuning (AT-RAFT) method, a novel LLM-based approach specifically designed for science policy interpretation. AT-RAFT incorporates three complementary artifacts: a two-stage retrieval mechanism, adaptive hard-negative fine-tuning, and an interpretable response interface. It is trained directly on policy documents, allowing the model to provide reference answers based on retrieved text while also offering the original policy context to enhance interpretability. Our experiments demonstrate that AT-RAFT improves retrieval accuracy by 48% and generation performance by 44% compared to existing baseline systems, effectively supporting real-world decision-making tasks for stakeholders in research institutions and funding agencies. Our proposed method has been adopted by <em>ScholarMate</em>, the largest professional research social networking platform in China, and is now deployed on their platform, providing global users with access to advanced policy interpretation tools. Additionally, a demo version of the instantiated interface is available at <span><span>https://github.com/renruntao/ResearchPolicy_RAG</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"278 ","pages":"Article 127330"},"PeriodicalIF":7.5,"publicationDate":"2025-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143748449","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A novel two-stage fuzzy classification method with different weight permutations for optimal GIS-based placement of wellness and sports centers","authors":"Behnam Razavian ,&nbsp;Seyed Masoud Hamed Seyedbeiglou ,&nbsp;Erfan Babaee Tirkolaee ,&nbsp;Ferzat Anka","doi":"10.1016/j.eswa.2025.127282","DOIUrl":"10.1016/j.eswa.2025.127282","url":null,"abstract":"<div><div>The optimal placement of wellness and sports centers is critical to maximizing their accessibility, effectiveness, and impact on public health. Strategic location planning ensures that these facilities are conveniently accessible to the largest possible segment of the population, thereby encouraging higher participation rates. Accessibility is particularly crucial in urban areas where space is limited, and in rural or underserved regions where health and recreational services are often scarce. Moreover, the strategic placement of these centers can enhance community cohesion and stimulate local economies. This study develops a novel sorting Multi-Criteria Decision-Making (MCDM) method called <em>fuzzy EDAS-Sort,</em> a variant of the Evaluation based on Distance from Average Solution (EDAS) ranking method through a fuzzy sorting with different weight permutations to address the optimal placement of wellness and sports centers through assigning alternatives to predefined and ordered classes. It aims to identify the best locations for wellness and sports centers in Ardabil, Iran by employing the fuzzy EDAS-Sort method which is the main contribution of this research combined with Geographic Information Systems (GIS). By integrating fuzzy set theory with EDAS-Sort and GIS, the inherent uncertainties are handled in performance evaluation and spatial data analysis. According to the findings, the fuzzy EDAS-Sort is computationally efficient and provide highly accurate classification results for the optimal placement of wellness and sports centers. Numerical results demonstrate that 20% of the studied locations belonged to the “Excellent and optimal area” class, 33.3% to the “Good area” class, and 53.3% to the “Above average area” class. Finally, sensitivity analysis reveals that the proposed method is stable against weight variations, with less than 2.78% fluctuation in the classification results, ensuring a high degree of robustness.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"278 ","pages":"Article 127282"},"PeriodicalIF":7.5,"publicationDate":"2025-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143748359","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Research on constructing and reasoning the collision knowledge graph of autonomous navigation ship based on enhanced BERT model","authors":"Zizhao Zhang ,&nbsp;Xinyue Yang ,&nbsp;Liping Sun ,&nbsp;Yu Sun ,&nbsp;Jichuan Kang","doi":"10.1016/j.eswa.2025.127429","DOIUrl":"10.1016/j.eswa.2025.127429","url":null,"abstract":"<div><div>This study proposes a Lexicon-enhanced BERT model (LEBERT) to enhance the analysis of autonomous ship collisions, addressing the challenges posed by limited annotated data and imbalanced entity type distribution. The LEBERT model integrates character-level and lexicon-level semantics through a Lexicon Adapter module and employs a span-based decoding approach to supplant traditional Conditional Random Fields (CRF). The experimental findings, drawn from an analysis of 257 accident reports, demonstrate that the proposed model attains an F1-score of 0.711, signifying enhancements of 2.3% and 21.3% over baseline models in the domains of general and long-entity recognition, respectively. The impact of entity label granularity on model performance was further explored, confirming that enhanced granularity greatly improves prediction accuracy and clustering consistency. Utilizing the entity prediction results derived from 60 accident cases, a novel conversion framework was developed that translates human error factors identified in conventional maritime analyses into failure modes applicable to autonomous ship systems. Through the selection of causal factors and the reconfiguration of case labels, a specialized knowledge graph tailored for autonomous navigation was developed. The knowledge graph incorporates advanced BERT-based reasoning within a multimodal data framework, thereby enabling the extraction of entities from unstructured text and supporting the development of scalable domain knowledge repositories. This study proposes a series of concepts relating to the implementation of knowledge graphs in the domain of autonomous navigation ships.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"278 ","pages":"Article 127429"},"PeriodicalIF":7.5,"publicationDate":"2025-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143748561","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"PATSA-BIL: Pipeline for automated texture and structure analysis of borehole image logs","authors":"André M. Souza ,&nbsp;Matheus A. Cruz ,&nbsp;Paola M.C. Braga ,&nbsp;Rodrigo B. Piva ,&nbsp;Rodrigo A.C. Dias ,&nbsp;Paulo R. Siqueira ,&nbsp;Willian A. Trevizan ,&nbsp;Candida M. de Jesus ,&nbsp;Camilla Bazzarella ,&nbsp;Rodrigo S. Monteiro ,&nbsp;Flavia C. Bernardini ,&nbsp;Leandro A.F. Fernandes ,&nbsp;Elaine P.M. de Sousa ,&nbsp;Daniel de Oliveira ,&nbsp;Marcos Bedo","doi":"10.1016/j.eswa.2025.127345","DOIUrl":"10.1016/j.eswa.2025.127345","url":null,"abstract":"<div><div>High-resolution image logs are essential for understanding rock compositions and geological features, but their analysis requires expertise and can be tedious and error-prone. This manuscript presents <span>PATSA-BIL</span>, a novel data-driven pipeline that uses Artificial Intelligence (AI) and computer vision techniques to enhance the analysis of resistivity image log data in the oil industry. <span>PATSA-BIL</span> enables deep-learning-based classification of resistivity textures and automatic segmentation of geological structures in borehole images. In addition to texture classification, <span>PATSA-BIL</span> provides of a series of chained Machine Learning (ML) tasks that comprise Simple Linear Iterative Clustering (SLIC) superpixels, radiomics embeddings, and Density-Based Spatial Clustering of Applications with Noise (DBSCAN) for geological structure segmentation. Experiments with the proposed pipeline, including comparisons against five fine-tuned, deep-learning models (Mask R-CNN, Faster R-CNN, RetinaNet, DETR, and DINOv2), demonstrate that <span>PATSA-BIL</span> achieves up to 90% accuracy in detecting resistivity textures. Furthermore, <span>PATSA-BIL</span>’s segmentation chain outperforms the closest competitor (DINOv2) by achieving a 16% higher Mean Average Precision (mAP), significantly improving the clustering of borehole structures by up to 7% while reducing computational demands compared to other approaches. Statistical validation further confirms the robustness of the proposed approach, performed using a Friedman test (<span><math><mi>p</mi></math></span>-value = <span><math><mrow><mn>6</mn><mo>.</mo><mn>7</mn><mi>⋅</mi><mn>1</mn><msup><mrow><mn>0</mn></mrow><mrow><mo>−</mo><mn>23</mn></mrow></msup></mrow></math></span>, <em>vs.</em> DINOv2) and a Wilcoxon–Holm post-hoc comparison between the top models (<span><math><mi>p</mi></math></span>-value = <span><math><mrow><mn>1</mn><mo>.</mo><mn>4</mn><mi>⋅</mi><mn>1</mn><msup><mrow><mn>0</mn></mrow><mrow><mo>−</mo><mn>3</mn></mrow></msup></mrow></math></span>, <em>vs.</em> DINOv2) with very large effect size.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"278 ","pages":"Article 127345"},"PeriodicalIF":7.5,"publicationDate":"2025-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143748557","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Short- and medium-term power load forecasting model based on a hybrid attention mechanism in the time and frequency domains","authors":"Ziran Peng,&nbsp;Xiaoyang Yang","doi":"10.1016/j.eswa.2025.127329","DOIUrl":"10.1016/j.eswa.2025.127329","url":null,"abstract":"<div><div>To address the challenges of high computational complexity, suboptimal feature selection, and information redundancy introduced by external influences in short- and medium-term power load forecasting, this paper proposes a novel short- and medium-term power load forecasting model, LDTformer, which incorporates a hybrid attention mechanism in the time and frequency domains and builds upon the Transformer architecture. The model mainly comprises the block sparse discrete cosine transform (BSDCT) module, seasonal-trend decomposition (STD) module, causal convolutional layer, and lagged correlation frequency domain attention (LCFA) mechanism. First, feature selection is optimized using Pearson correlation to reduce the interference of non-essential information. Then, the BSDCT module is employed to map the data to the frequency domain, improving the accuracy and efficiency of load feature extraction while reducing the computational complexity of the model. Moreover, the input features are decomposed into trend and seasonal components through the STD module, which comprises multiple expert and gated networks, to improve the forecasting accuracy and model adaptability. Finally, the LCFA mechanism introduces learnable lag weights to enhance the model’s ability to globally capture the periodicity and trends of the time series, achieving complementarity between the frequency and time domain features, thus improving the accuracy and efficiency of predictions. The experimental results show that LDTformer significantly outperforms the comparison models Transformer, FEDformer, Mamba, and LSTM in several evaluation metrics. Its predictions are highly consistent with the actual values and demonstrate excellent robustness and stability across different seasons and sampling points.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"278 ","pages":"Article 127329"},"PeriodicalIF":7.5,"publicationDate":"2025-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143748633","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multi-objective optimization of software testing schedules for modular control Software, considering learning and negligence factors of testing Staffs","authors":"Chih-Chiang Fang ,&nbsp;Chun-Wu Yeh","doi":"10.1016/j.eswa.2025.127359","DOIUrl":"10.1016/j.eswa.2025.127359","url":null,"abstract":"<div><div>Comprehensive software testing and debugging are essential to prevent customer dissatisfaction resulting from software defects. However, software development managers often encounter significant constraints, such as limited budgets and time-sensitive market opportunities, which hinder extensive debugging efforts. Effective project management in software testing requires a careful balance of cost, quality, and timeline considerations to facilitate optimal decision-making. Traditional Software Reliability Growth Models (SRGMs) typically focus on single-system testing; however, modern control software for machine or robot development increasingly adopts a modular systems engineering approach. This transition allows for parallel testing across multiple modules, enabling managers to assign different testing teams to work concurrently on separate components. This study addresses the need for a structured approach by proposing multi-objective programming models to optimize testing schedules for complex, modular software systems. These models aim to minimize overall testing time while ensuring that reliability standards are met within resource constraints. Furthermore, this study introduces a novel SRGM that incorporates the learning curves of testing personnel and potential negligence factors. The fitting results for prediction accuracy can reach 95% in the majority of the cases analyzed in this study. Additionally, we propose the development of a computerized decision support system to facilitate practical implementation and decision-making in real-world scenarios.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"279 ","pages":"Article 127359"},"PeriodicalIF":7.5,"publicationDate":"2025-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143760701","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Gwo-ga-xgboost-based model for Radio-Frequency power amplifier under different temperatures","authors":"Shaohua Zhou","doi":"10.1016/j.eswa.2025.127439","DOIUrl":"10.1016/j.eswa.2025.127439","url":null,"abstract":"<div><div>To improve the modeling accuracy and modeling speed of the XGBoost model, a Gray Wolf Optimization (GWO)-Genetic Algorithm (GA)-XGBoost model is proposed in this paper and is applied to model radio frequency (RF) power amplifiers at different temperatures. The experimental results show that compared to XGBoost, GA-XGBoost, and GWO-XGBoost, the modeling accuracy of GWO-GA-XGBoost can be improved by one order of magnitude or more. Compared to XGBoost, GA-XGBoost, and GWO-XGBoost, GWO-GA-XGBoost has also increased the modeling speed by one magnitude or more. In addition, compared to the classic machine learning algorithms gradient boosting, random forest, and AdaBoost, the proposed GWO-GA-XGBoost can improve modeling accuracy by two or more orders of magnitude while also increasing modeling speed by one or more.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"278 ","pages":"Article 127439"},"PeriodicalIF":7.5,"publicationDate":"2025-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143734877","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Cluster-Boosted Artificial Neural Networks: Theory, implementation, and performance evaluation","authors":"George Papazafeiropoulos","doi":"10.1016/j.eswa.2025.127332","DOIUrl":"10.1016/j.eswa.2025.127332","url":null,"abstract":"<div><div>This study introduces a new clustering technique to boost Artificial Neural Networks’ (ANNs’) performance. The term “Cluster-Boosted Artificial Neural Networks” (“CBANNs”) is coined for ANNs using this technique. By adding cluster identifiers as extra input features, CBANNs enhance conventional ANNs and improve the model’s ability to identify underlying patterns in complicated data landscapes. This method offers a solution to some limitations of standard ANNs, which often struggle with high-dimensional data, local minima, and nonlinear relationships. Without the need for manual feature engineering or in-depth domain knowledge, CBANNs greatly increase prediction accuracy by employing unsupervised clustering, using k-medoids, to build a more structured input space. Various numerical results are presented which validate the superior predictive ability of CBANNs across nine benchmark functions, including De Jong’s 5th, Griewank, and Rastrigin functions. Compared to conventional ANNs with identical hyperparameters, CBANNs achieve error reductions of up to 98%, consistently demonstrating higher performance on functions with intricate geometries and multiple minima. Furthermore, CBANNs are applied to a terrain modeling problem, which proved that CBANNs can reduce the prediction error by up to 95% compared to standard ANNs, indicating their potential for high-precision applications. These findings underscore the CBANN’s ability to generalize effectively in challenging datasets, suggesting its broader applicability in fields that demand accuracy in the presence of complex data distributions.</div></div>","PeriodicalId":50461,"journal":{"name":"Expert Systems with Applications","volume":"278 ","pages":"Article 127332"},"PeriodicalIF":7.5,"publicationDate":"2025-03-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143724872","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}