检索结果-内蒙古大学图书馆

arXiv 2023年

作者： Alhindi, Tariq Chakrabarty, Tuhin Musi, Elena Muresan, Smaranda Department of Computer Science Columbia University United States Data Science Institute Columbia University United States Department of Communication and Media University of Liverpool United Kingdom

Fallacies are used as seemingly valid arguments to support a position and persuade the audience about its validity. Recognizing fallacies is an intrinsically difficult task both for humans and machines. Moreover, a big challenge for computational models lies in the fact that fallacies are formulated differently across the datasets with differences in the input format (e.g., question-answer pair, sentence with fallacy fragment), genre (e.g., social media, dialogue, news), as well as types and number of fallacies (from 5 to 18 types per dataset). To move towards solving the fallacy recognition task, we approach these differences across datasets as multiple tasks and show how instruction-based prompting in a multitask setup based on the T5 model improves the results against approaches built for a specific dataset such as T5, BERT or GPT-3. We show the ability of this multitask prompting approach to recognize 28 unique fallacies across domains and genres and study the effect of model size and prompt choice by analyzing the per-class (i.e., fallacy type) results. Finally, we analyze the effect of annotation quality on model performance, and the feasibility of complementing this approach with external knowledge. © 2023, CC BY.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Interpretable Skill Learning for Dynamic Treatment Regimes through Imitation

Interpretable Skill Learning for Dynamic Treatment Regimes t...

引用

Annual Conference on Information sciences and Systems (CISS)

作者： Yushan Jiang Wenchao Yu Dongjin Song Wei Cheng Haifeng Chen Department of Computer Science and Engineering University of Connecticut Storrs CT USA NEC Laboratories American Data Science & System Security Princeton NJ USA

Imitation learning that mimics experts' skills from their demonstrations has shown great success in discovering dynamic treatment regimes, i.e., the optimal decision rules to treat an individual patient based on related evolving treatment and covariate history. Existing imitation learning methods, however, still lack the capability to interpret the underlying rationales of the learned policy in a faithful way. Moreover, since dynamic treatment regimes for patients often exhibit varying patterns, i.e., symptoms that transit from one to another, the flat policy learned by a vanilla imitation learning method is typically undesired. To this end, we propose an Interpretable Skill Learning (ISL) framework to resolve the aforementioned challenges for dynamic treatment regimes through imitation. The key idea is to model each segment of experts' demonstrations with a prototype layer and integrate it with the imitation learning layer to enhance the interpretation capability. On one hand, the ISL framework is able to provide interpretable explanations by matching the prototype to exemplar segments during the inference stage, which enables doctors to perform reasoning of the learned demonstrations based on human-understandable patient symptoms and lab results. On the other hand, the obtained skill embedding consisting of prototypes serves as conditional information to the imitation learning layer, which implicitly guides the policy network to provide a more accurate demonstration when the patients' state switches from one stage to another. Thoroughly empirical studies demonstrate that our proposed ISL technique can achieve better performance than state-of-the-art methods. Moreover, the proposed ISL framework also exhibits good interpretability which cannot be observed in existing methods.

关键词： Learning systems Visualization MIMICs Decision making Prototypes Medical services Cognition

来源：评论

学校读者我要写书评

暂无评论

How deep learning influences workflows and roles in virtual surgical planning

引用

Discover Health Systems 2023年第1期2卷 1-17页

作者： Hofer, Beat Kittler, Markus Laukens, Kris Optealis Obfelden Switzerland Department of Management & Law MCI|The Entrepreneurial School Innsbruck Austria Adrem Data Lab Department of Computer Science University of Antwerp Antwerp Belgium

Deep learning (DL) has the potential to transform surgical practice, altering workflows and changing the roles of practitioners involved. However, studies have shown that introducing such change requires user acceptance. Following the development and presentation of a visual prototype for planning facial surgery interventions, the project aimed to understand the utility of DL, the implied workflow and role changes it would entail, and the potential barriers to its adoption in practice. This paper presents a multi-year case study providing insights from developing and introducing a visual prototype. The prototype was co-developed by facial surgeons, DL experts, and business process engineers. The study uses project data involving semi-structured interviews, workgroup results, and feedback from an external practitioner audience exposed to the prototype regarding their views on adopting DL tools in practice. The surgeons attested a high utility to the application. However, the data also highlights a perceived need to remain in control, be able to intervene, and override surgical workflows in short intervals. Longer intervals without opportunities to intervene were seen with skepticism, suggesting that the practitioners’ acceptance of DL requires a carefully designed workflow in which humans can still take control of events. Deep learning can improve and accelerate facial surgery intervention planning. Models from the business and management literature partially explain the acceptance of new technologies. Perceived ease of use seems less relevant than the perceived usefulness of new technology. Involving algorithms in clinical decision-making will change workflows and professional identities.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Sentiment Analysis on Drug Reviews Using Diverse Classification Techniques 3

Sentiment Analysis on Drug Reviews Using Diverse Classificat...

引用

3rd International Conference on Communication, Computing and Industry 4.0, C2I4 2022

作者： Duraisamy, Premkumar Natarajan, Yuvaraj Preethaa, K.R.Sri Mouthami, K. KPR Institute of Engineering and Technology Department of Artificial Intelligence and Data Science Coimbatore India KPR Institute of Engineering and Technology Department of Computer Science and Engineering Coimbatore India Research Professor Intelligent Construction Automation Center Kyungpook National University Deagu Korea Republic of

ISBN: (纸本)9781665474979

Free text reviews are abundantly distributed over the internet among the standard population of the world. Although online doctors were consulted during the pandemic, many drugs are taken without prescription by ordinary people. Reviews are distributed about the medicines through online feeds, posts and blogs, which are merely helpful in identifying the side effects and their adverse reactions in combination with another drug or standalone adverse reactions. Spelling mistakes, jargon and short names are normalized throughout the medical terminologies using Natural language processing (NLP) and RxNorm from the reviews extracted from these sources. The concept of Bidirectional Encoder Representations from Transformers (BERT) is used to mine the association rules for the drug interactions when used in conjugation with another drug, and the idea of Adaptive Fuzzy logic neural networks is used to validate the results of association rules derived from the BERT. Inconsistencies identified by the regulations generated were rejected efficiently by the Support vector machines (SVM), Radial Basis Function (RBF)and the system shows 60.11 percentage of accuracy with 10 cross-fold validation, whereas the results are equivalent to the accuracy obtained using SVM with RBF kernel. The data set is reduced from its dimension using Linear discriminant analysis (LDA) by applying the mining process and BERT on two drugs, namely 'Azithromycin' in conjugation with Buprenorphine has Type 'A' side effects, and Prozac in conjugation with Metronidazole shows different side effects. Deplin and Zoloft give very positive symptoms and are independent of dosage interactions. © 2022 IEEE.

关键词： Bidirectional Encoder Representations from Transformers (BERT) Linear discriminant analysis (LDA) Radial Basis Function (RBF) Support vector machines (SVM)

来源：评论

学校读者我要写书评

暂无评论

Evaluation of CAN Bus Security Vulnerabilities and Potential Solutions

Evaluation of CAN Bus Security Vulnerabilities and Potential...

引用

Women in data science at Prince Sultan University (WiDS PSU), International Conference of

作者： Asma Alfardus Danda B. Rawat Department of Electrical Engineering and Computer Science Data Science and Cybersecurity Center (DSC2) Howard University Washington D.C USA

The past decade has seen a significant increase in the automobile industry, which has come with some serious challenges and threats. Modem vehicles are now made up of complex mechanical systems, as well as sophisticated electronic devices and connections to the outside world. Various electronic devices utilize standard communication protocols, including the Controller Area Network (CAN), to establish communication with each other. Unfortunately, CAN lacks some fundamental security features, such as encryption and authentication, which makes it vulnerable to security attacks. This can lead to accidents and financial losses for the users of these vehicles. To address this issue, researchers have proposed a number of security measures, such as cryptography and Intrusion Detection Systems (IDS). This paper addresses the security vulnerabilities associated with CAN and proposes potential solutions to overcome its limitations.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Phishing Web Page Detection using Web Scraping

Phishing Web Page Detection using Web Scraping

引用

IEEE Southeastcon

作者： Mallika Boyapati Ramazan Aygun School of Data Science and Analytics Kennesaw State University Kennesaw GA USA Department of Computer Science Kennesaw State University Kennesaw GA USA

In today’s Internet era, webpages act as major user interfaces for the applications hosted by the organizations. One of the most common security attacks on the web page applications is phishing attacks. Phishing is a social engineering attack in which an adversary tries to steal user credentials by tricking them to believe that they are on a legitimate web page. Adversaries are using sophisticated and new ways to forge the web page designs craftily to trick the users into visiting the malicious links. The phishing webpages are used as a medium to carry out the art of phishing attacks. Web scraping is a methodology to extract features of each webpage. In this paper, web scraping is employed to extract hybrid feature set to implement Machine Learning (ML) models. The machine learning models like XGBoost, Multilayer Perceptron, Logistic Regression, SVM, Auto Encoder, Random Forest, Decision Trees, K-means, and Naive Bayes are built. All the models are tested with and without applying principal component analysis (PCA), a feature reduction technique. Extracting hybrid features by employing web scraping to train the ML models and finding the key features contributing to phishing detection on Phishpedia, Kaggle, and PhishTank datasets is the key contribution of this paper. Results show that XGBoost algorithm outperformed the all other classifiers with 98% accuracy or higher on the web scraped features on all the datasets. The model achieved higher precision and recall when compared to other approaches like CANTINA+, URL based approaches, SenseInput, Knowing Thy Domain, and Phishpedia on the Phishpedia dataset.

关键词： Uniform resource locators Visualization Machine learning algorithms Web page design Phishing Virtual assistants Web pages

来源：评论

学校读者我要写书评

暂无评论

A churn prediction dataset from the telecom sector: a new benchmark for uplift modeling

arXiv

引用

arXiv 2023年

作者： Verhelst, Théo Mercier, Denis Shrestha, Jeevan Bontempi, Gianluca Machine Learning Group Department of Computer Science Université Libre de Bruxelles Brussels Belgium Data Science Team Orange Belgium Brussels Belgium

Uplift modeling, also known as individual treatment effect (ITE) estimation, is an important approach for data-driven decision making that aims to identify the causal impact of an intervention on individuals. This paper introduces a new benchmark dataset for uplift modeling focused on churn prediction, coming from a telecom company in Belgium, Orange Belgium. Churn, in this context, refers to customers terminating their subscription to the telecom service. This is the first publicly available dataset offering the possibility to evaluate the efficiency of uplift modeling on the churn prediction problem. Moreover, its unique characteristics make it more challenging than the few other public uplift datasets. © 2023, CC BY-NC-ND.

关键词： Decision making

来源：评论

学校读者我要写书评

暂无评论

Early Identification of Cancer Blood Disorder Using Deep Convolutional Neural Networks 23

Early Identification of Cancer Blood Disorder Using Deep Con...

引用

Proceedings of the 5th International Conference on Information Management & Machine Intelligence

作者： Pulla Sujarani M. Yogeshwari K. Kalaiselvi Research Scholar Department of Computer Science Vels Institute of Science Technology & Advanced Studies (VISTAS) India Department of Information Technology Vels Institute of Science Technology and Advanced Studies(VISTAS) India Department of Data Analytics Saveetha College of Liberal Arts and Sciences SIMATS India

ISBN: (纸本)9798400709418

A cancer blood disorder can be dangerous if not detected in time. It causes aberrant white blood cell production in the blood by the bone marrow. Using image processing of microscopic images of the blood, it may be quickly diagnosed. Deep learning techniques are a practical approach to cancer blood disorders in early diagnosis. In this study, we have proposed a novel method to identify cancer blood disorders in the early stage using the deep convolutional neural network (DCNN). Using filtering techniques, the microscopic images are first preprocessed. The 2D Adaptive Anisotropic Diffusion Filter (2DAADF) technique is used for image filtering to eliminate noise from the input images. Feature extraction is carried out utilizing the Grey Level Co-Occurrence Matrix (GLCM) from the filtered images to increase the identification accuracy. Finally, the proposed DCNN classifier is used to detect the cancer blood disorder. In comparison to the current approaches, the proposed methodology attained the maximum accuracy of 97%. According to the results, the proposed method can identify blood cancer with high accuracy and may help with its diagnosis and treatment in the early stages.

关键词： Cancer Blood Disorder 1

来源：评论

学校读者我要写书评

暂无评论

The analysis of TikTok posts of Malaysian influencers: post categorization and strategies to improve popularity

The analysis of TikTok posts of Malaysian influencers: post ...

引用

International Conference on Green Energy, Computing and Intelligent Technology (GEn-CITy 2023)

作者： K. R. Purba R. Ramli Data Science Western Digital KL Center of Excellence Selangor Malaysia Department of Computer Science University of Southampton Malaysia Iskandar Puteri Johor Malaysia

Gaining popularity on social media is important for many users, especially on platforms like TikTok where people can earn revenue. Various impactful studies have been conducted on popularity analysis and user profiling on many social networks. However, studies on TikTok were focused on thematic analysis, such as on health, creativity, and awareness. Thus, this paper aims to categorize (cluster) TikTok posts and identify the factors that contribute to popularity, using Kmeans clustering. The data was collected from the posts of Malaysian influencers over six months period. Three metrics were used, namely exposure, attraction and virality. The posts were categorized into casual, quality, volume, supporting, and influencer posts. Important insights were acquired from the analysis, such as the best day/time to post, the ideal numbers for video duration, following, biography length, caption length, and hashtags, including the types of hashtags to use. This study will help TikTok users as well as brand owners for marketing decisions.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Feature Selection in Associative Classification-A Review and Comparative Analysis

Feature Selection in Associative Classification-A Review and...

引用

Information Technology (ACIT)

作者： Raed Alazaidah Ahmad Al-qerem Mais Haj Qasem Ala'a Al-shaikh Nabeel Almilli MohammadNoor Injadat Department of Data Science and AI Faculty of Information Technology Zarqa University Zarqa Jordan Department of Computer Science Faculty of Information Technology Zarqa University Zarqa Jordan

Feature selection is one of the most important and significant steps that highly affects the final accuracy of any classification model. It aims to determine the most significant features that help in accurately predicting the class label for any new case. This research aims to survey and identify the best feature selection method that suits algorithms belong to Associative Classification (AC). Hence, nine different feature selection methods have been evaluated and compared using five AC based algorithms with respect to Accuracy metric. Results revealed that ABB-IEP and ABB-LIU methods show the best performance with datasets that contain nominal features only, while Relief feature selection method showed the best Accuracy with datasets that comprise Real and Integer features.

关键词：

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：