Machine learning with applications最新文献

Deep learning models for enhanced in-field maize leaf disease diagnosis 增强田间玉米叶片病害诊断的深度学习模型

Machine learning with applications Pub Date : 2025-05-26 DOI: 10.1016/j.mlwa.2025.100673

Joyce Nakatumba-Nabende , Sudi Murindanyi

{"title":"Deep learning models for enhanced in-field maize leaf disease diagnosis","authors":"Joyce Nakatumba-Nabende , Sudi Murindanyi","doi":"10.1016/j.mlwa.2025.100673","DOIUrl":"10.1016/j.mlwa.2025.100673","url":null,"abstract":"<div><div>Maize leaf diseases significantly threaten crop yields, and there is need for accurate, and accessible diagnostic tools. This research addresses this need by developing and evaluating deep learning (DL) and machine learning (ML) models for in-field classification and detection of four critical maize diseases: Maize Leaf Blight, Maize Lethal Necrosis, Maize Streak Virus, and Fall Armyworm damage. Utilizing field imagery captured via digital cameras and smartphones across Uganda, Tanzania, Ghana, and Namibia, we developed and compared custom Convolutional Neural Networks (CNNs), transfer learning (MobileNetV2, InceptionResNetV2), Vision Transformers (ViT), and classical ML models. For detection, a transformer-enhanced YOLOv10 architecture was implemented. Explainable AI (XAI) techniques (Grad-CAM, LIME) were incorporated to ensure model transparency. MobileNetV2 achieved the highest classification accuracy (97%), while the enhanced YOLOv10 reached a mean Average Precision (mAP) of 0.995 for object detection. The best models were integrated into a mobile application deployed on edge devices for real-time diagnosis by smallholder farmers in Uganda, with performance validated through field tests. This study demonstrates a powerful, interpretable, and field-deployable solution combining advanced DL, transformers, and XAI for effective maize health management.</div></div>","PeriodicalId":74093,"journal":{"name":"Machine learning with applications","volume":"20 ","pages":"Article 100673"},"PeriodicalIF":0.0,"publicationDate":"2025-05-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144138015","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A novel unsupervised fine-tuning method for text summarization, and highlighting the limitations of ROUGE score 一种新的文本摘要的无监督微调方法，突出了ROUGE评分的局限性

Machine learning with applications Pub Date : 2025-05-22 DOI: 10.1016/j.mlwa.2025.100666

Ala Alam Falaki, Robin Gras

{"title":"A novel unsupervised fine-tuning method for text summarization, and highlighting the limitations of ROUGE score","authors":"Ala Alam Falaki, Robin Gras","doi":"10.1016/j.mlwa.2025.100666","DOIUrl":"10.1016/j.mlwa.2025.100666","url":null,"abstract":"<div><div>The limited availability of datasets for text summarization tasks and their similar characteristics (e.g. news articles) make it crucial to focus on unsupervised learning techniques to enable summarization across different domains. Moreover, since summarization produces text output, effective methods developed for news articles can be applied to other domains lacking sufficient labeled data. This study introduces a novel target selection process to be used as an unsupervised learning method for fine-tuning text summarization models with unlabeled data. The process involves two-steps: first, generating an extractive summary (Ext-Reference) from the article, and second, using an abstractive model to create a pool of candidate summaries. The most suitable summary (to be used as the target) is then selected by calculating the cosine similarity between the Ext-Reference’s embedding and each candidate’s embedding. Furthermore, this project underscores the limitations of the ROUGE score, which assigns a relatively low score to this method. However, extended analysis with various metrics, including using GPT-4 as a judge, demonstrates the effectiveness of this technique for fine-tuning models without a specific target reference. It highlights the importance of using a combination of metrics, like those included in the SumEvaluator package released alongside this paper. SumEvaluator package on Github: <span><span>https://github.com/AlaFalaki/SumEvaluator</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":74093,"journal":{"name":"Machine learning with applications","volume":"20 ","pages":"Article 100666"},"PeriodicalIF":0.0,"publicationDate":"2025-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144167464","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Neural Network-Aided NILM (NNAN) disaggregation: Revealing appliance consumption patterns with iterative subtraction 神经网络辅助NILM （NNAN）分解：用迭代减法揭示家电消费模式

Machine learning with applications Pub Date : 2025-05-22 DOI: 10.1016/j.mlwa.2025.100667

Yacine Belguermi, Patrice Wira, Gilles Hermann

{"title":"Neural Network-Aided NILM (NNAN) disaggregation: Revealing appliance consumption patterns with iterative subtraction","authors":"Yacine Belguermi, Patrice Wira, Gilles Hermann","doi":"10.1016/j.mlwa.2025.100667","DOIUrl":"10.1016/j.mlwa.2025.100667","url":null,"abstract":"<div><div>Non-Intrusive Load Monitoring (NILM) is a method to decompose overall electricity consumption into individual appliance-level data, utilizing the primary meter’s readings without additional sensors on each device. This article introduces a novel approach which is a Neural Network-Aided NILM (NNAN), focusing on revealing appliance consumption patterns by following a sequential subtraction method. Our goal is to tackle the issue where high-power and highly-used appliances make it difficult for neural networks to accurately separate the usage of lower-power and less-used appliances. We mainly employ Long Short-Term Memory (LSTM) and Convolutional Neural Networks (CNN) using inception blocks as key components. Our proposed architecture is validated on three public datasets that are AMPds2, ECO and UK-DALE. The NNAN model showed promising results, achieving disaggregation accuracy improvements of up to 5.13% on AMPds2, 3.79% on ECO, and 9.55% on UK-DALE compared to the reference methods. Additionally, NNAN reduces model complexity, requiring up to 74% fewer parameters than traditional deep learning approaches, leading to improved computational efficiency. Finally, NNAN demonstrated a reduced correlation between appliance usage rates and disaggregation accuracies.</div></div>","PeriodicalId":74093,"journal":{"name":"Machine learning with applications","volume":"20 ","pages":"Article 100667"},"PeriodicalIF":0.0,"publicationDate":"2025-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144135085","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Bitcoin price direction prediction using on-chain data and feature selection 利用链上数据和特征选择进行比特币价格方向预测

Machine learning with applications Pub Date : 2025-05-20 DOI: 10.1016/j.mlwa.2025.100674

Ritwik Dubey , David Enke

{"title":"Bitcoin price direction prediction using on-chain data and feature selection","authors":"Ritwik Dubey , David Enke","doi":"10.1016/j.mlwa.2025.100674","DOIUrl":"10.1016/j.mlwa.2025.100674","url":null,"abstract":"<div><div>Bitcoin is the most traded cryptocurrency by volume and market cap. A number of scholars have directed their research towards characterizing Bitcoin’s speculative behavior using a myriad of techniques such as technical analysis, price regression, and direction classification. For this work, research is conducted using the relatively nascent technique of on-chain data analysis. The goal of this research is to evaluate Bitcoin’s on-chain data in predicting future price direction. First, a classification process of on-chain data features that helps the reader understand their relevance is proposed. To address the curse of dimensionality, feature selection algorithms such as L1 regression, Boruta, and the dimensionality reduction algorithm Principal Component Analysis (PCA) are utilized. The research then explores advanced neural networks for next day price direction prediction, including the Convolutional Neural Network-Long-Short Term Memory (CNN-LSTM) and the Temporal Convolutional Network (TCN). Neural network models and trading strategies are then compared based on their return statistics. A comparative analysis of feature selection, learning model performance, and trading strategy performance is also conducted. Results from the research show that the Boruta feature selection algorithm combined with the CNN-LSTM model performs best compared to other combinations with a prediction accuracy of 82.03 % over the testing period. In addition, the on-chain features within the category, realized value, and unrealized value classifications have higher predictive powers for next day price direction prediction. Finally, during trade simulations, the CNN-LSTM model with a Long-Short strategy had an annualized return of 1682.7 % and a Sharpe Ratio of 6.47.</div></div>","PeriodicalId":74093,"journal":{"name":"Machine learning with applications","volume":"20 ","pages":"Article 100674"},"PeriodicalIF":0.0,"publicationDate":"2025-05-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144167463","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Transformers—Messages in disguise 变形金刚——伪装的信息

Machine learning with applications Pub Date : 2025-05-19 DOI: 10.1016/j.mlwa.2025.100669

Joshua H. Tyler , Donald R. Reising , Mohamed M.K. Fadul

引用次数: 0

LARE: Latent augmentation using regional embedding with vision-language model LARE：基于区域嵌入的视觉语言模型的潜在增强

Machine learning with applications Pub Date : 2025-05-19 DOI: 10.1016/j.mlwa.2025.100671

Kosuke Sakurai , Tatsuya Ishii , Ryotaro Shimizu , Linxin Song , Masayuki Goto

{"title":"LARE: Latent augmentation using regional embedding with vision-language model","authors":"Kosuke Sakurai , Tatsuya Ishii , Ryotaro Shimizu , Linxin Song , Masayuki Goto","doi":"10.1016/j.mlwa.2025.100671","DOIUrl":"10.1016/j.mlwa.2025.100671","url":null,"abstract":"<div><div>In recent years, considerable research has been conducted on vision-language models (VLMs) that handle both image and text data; these models are being applied to diverse downstream tasks, such as “image-related chat,” “image recognition by instruction,” and “answering visual questions.” Vision-language models, such as Contrastive Language–Image Pre-training (CLIP), are also high-performance image classifiers, and are being developed into domain adaptation methods that can utilize language information to extend into unseen domains. However, because these VLMs embed images as a single point in a unified embedding space, they do not fully exploit the diverse domain performance of large-scale vision-language models. Therefore, in this study, we proposed the <em>Latent Augmentation using Regional Embedding</em> (LARE), which embeds the image as a region in the unified embedding space learned by the VLM. By sampling the augmented image embeddings from within this latent region, LARE enables data augmentation to various unseen domains, not just to specific unseen domains. LARE achieves robust image classification for domains in and out using augmented image embeddings to fine-tune VLMs. We demonstrate that LARE outperforms previous fine-tuning models in terms of image classification accuracy on three benchmarks. We also demonstrate that LARE is a more robust and general model that is valid under multiple conditions, such as unseen domains, small amounts of data, and imbalanced data.</div></div>","PeriodicalId":74093,"journal":{"name":"Machine learning with applications","volume":"20 ","pages":"Article 100671"},"PeriodicalIF":0.0,"publicationDate":"2025-05-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144105598","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A credit card fraud detection approach based on ensemble machine learning classifier with hybrid data sampling 基于混合数据采样的集成机器学习分类器的信用卡欺诈检测方法

Machine learning with applications Pub Date : 2025-05-18 DOI: 10.1016/j.mlwa.2025.100675

Khanda Hassan Ahmed , Stefan Axelsson , Yuhong Li , Ali Makki Sagheer

引用次数: 0

“A comprehensive review of AI-enhanced decision making: An empirical analysis for optimizing medication market business” 《人工智能增强决策的综合评价：优化药品市场业务的实证分析》

Machine learning with applications Pub Date : 2025-05-18 DOI: 10.1016/j.mlwa.2025.100676

Zainab Nadhim Jawad, Dr. Villányi Balázs János

{"title":"“A comprehensive review of AI-enhanced decision making: An empirical analysis for optimizing medication market business”","authors":"Zainab Nadhim Jawad, Dr. Villányi Balázs János","doi":"10.1016/j.mlwa.2025.100676","DOIUrl":"10.1016/j.mlwa.2025.100676","url":null,"abstract":"<div><div>Enterprise Resource Planning (ERP) systems play a critical role in integrating key business functions, including customer relationship management (CRM), inventory control, and financial operations. The integration of Artificial Intelligence (AI) techniques, particularly Machine Learning (ML), has the potential to enhance decision-making and optimize operational efficiency. This study systematically reviews AI-driven enhancements in ERP using the PRISMA methodology to identify trends, applications, and challenges. Additionally, an empirical analysis using a publicly available dataset conducted to demonstrate the impact of ML-driven sentiment analysis on demand forecasting in the pharmaceutical sector. Our findings indicate that AI-enhanced ERP systems improve forecasting accuracy, inventory management, and financial planning, leading to better alignment with market demands. Further, empirical results highlight the transformative role of AI techniques in optimizing ERP functionalities and supporting data-driven decision-making. This research provides actionable insights for enterprises aiming to integrate ML techniques into ERP systems to enhance business performance.</div></div>","PeriodicalId":74093,"journal":{"name":"Machine learning with applications","volume":"20 ","pages":"Article 100676"},"PeriodicalIF":0.0,"publicationDate":"2025-05-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144124993","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Automatic and accurate measurement of cattle body based on a lightweight YOLOv8-Pose model and 3D point cloud

Machine learning with applications Pub Date : 2025-05-10 DOI: 10.1016/j.mlwa.2025.100662

Chunhong Yue , Xiaohang Zhang , Yajun Zhang , Haijun Jiang , Haoyuan Miao

{"title":"Automatic and accurate measurement of cattle body based on a lightweight YOLOv8-Pose model and 3D point cloud","authors":"Chunhong Yue , Xiaohang Zhang , Yajun Zhang , Haijun Jiang , Haoyuan Miao","doi":"10.1016/j.mlwa.2025.100662","DOIUrl":"10.1016/j.mlwa.2025.100662","url":null,"abstract":"<div><div>In animal husbandry, the transition from traditional manual and time-consuming methods to non-contact body measurements for cattle has become increasingly important. Current research predominantly relies on 3D data to extract body measurement key points; however, the presence of noise in the data often leads to significant errors. Furthermore, the complexity of deep learning models often results in poor real-time performance in practical applications. To address these challenges, this study presents the YOLOv8-Pose model with a lightweight shared convolutional detection head (YOLOv8-Pose-LSCDH), which integrates with 3D point clouds for the automatic measurement of dairy cattle. Initially, this method employs a Kinect V2 sensor to capture RGB and depth images of the cattle. The key points predicted by the YOLOv8-Pose-LSCDH on the RGB images are subsequently projected onto the 3D point cloud generated from the depth image, thereby obtaining three-dimensional body measurement key points. These key points are then used to compute the cattle's body dimensions, including withers height, hip height, chest depth, chest circumference, body length, and hip length. Experimental results demonstrate that, compared to the baseline model, the proposed YOLOv8-Pose-LSCDH model requires fewer parameters and less computational load without compromising accuracy, thus offering superior real-time performance. Additionally, when compared to manual measurement, the proposed body measurement method maintains an overall relative error of <4 %.</div></div>","PeriodicalId":74093,"journal":{"name":"Machine learning with applications","volume":"20 ","pages":"Article 100662"},"PeriodicalIF":0.0,"publicationDate":"2025-05-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143943547","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Classification with reject option: Distribution-free error guarantees via conformal prediction 带拒绝选项的分类：通过适形预测保证无分布错误

Machine learning with applications Pub Date : 2025-05-09 DOI: 10.1016/j.mlwa.2025.100664

Johan Hallberg Szabadváry , Tuwe Löfström , Ulf Johansson , Cecilia Sönströd , Ernst Ahlberg , Lars Carlsson

{"title":"Classification with reject option: Distribution-free error guarantees via conformal prediction","authors":"Johan Hallberg Szabadváry , Tuwe Löfström , Ulf Johansson , Cecilia Sönströd , Ernst Ahlberg , Lars Carlsson","doi":"10.1016/j.mlwa.2025.100664","DOIUrl":"10.1016/j.mlwa.2025.100664","url":null,"abstract":"<div><div>Machine learning (ML) models always make a prediction, even when they are likely to be wrong. This causes problems in practical applications, as we do not know if we should trust a prediction. ML with reject option addresses this issue by abstaining from making a prediction if it is likely to be incorrect. In this work, we formalise the approach to ML with reject option in binary classification, deriving theoretical guarantees on the resulting error rate. This is achieved through conformal prediction (CP), which produce prediction sets with distribution-free validity guarantees. In binary classification, CP can output prediction sets containing exactly one, two or no labels. By accepting only the singleton predictions, we turn CP into a binary classifier with reject option.</div><div>Here, CP is formally put in the framework of predicting with reject option. We state and prove the resulting error rate, and give finite sample estimates. Numerical examples provide illustrations of derived error rate through several different conformal prediction settings, ranging from full conformal prediction to offline batch inductive conformal prediction. The former has a direct link to sharp validity guarantees, whereas the latter is more fuzzy in terms of validity guarantees but can be used in practice. Error-reject curves illustrate the trade-off between error rate and reject rate, and can serve to aid a user to set an acceptable error rate or reject rate in practice.</div></div>","PeriodicalId":74093,"journal":{"name":"Machine learning with applications","volume":"20 ","pages":"Article 100664"},"PeriodicalIF":0.0,"publicationDate":"2025-05-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143935733","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0