Article
Mathematics, Applied
Kan Chen, Xufei Tang, Jiangfeng Hao
Summary: This paper investigates the complete consistency and convergence rate of the nearest neighbor estimator of the density function based on WOD samples using the exponential inequality of WOD random variables. The results generalize and improve some corresponding ones in the literature, with a weaker restriction on the dominating coefficients g(n). Even with the geometric growth of g(n), the consistency result and convergence rate still hold.
JOURNAL OF MATHEMATICAL INEQUALITIES
(2021)
Article
Automation & Control Systems
Laszlo Gyorfi, Roi Weiss
Summary: This study investigates the universal consistency and convergence rates of simple nearest-neighbor prototype rules for multiclass classification in metric spaces. The newly introduced Proto-NN rule simplifies the implementation and achieves consistent results without bounding assumptions on data distribution. Hybridizing between k-NN and Proto-NN, a second prototype rule achieves similar computational advantages as Proto-NN while maintaining the convergence rates of k-NN.
JOURNAL OF MACHINE LEARNING RESEARCH
(2021)
Article
Computer Science, Artificial Intelligence
Chunrong Wu, Qinglan Peng, Jia Lee, Kenji Leibnitz, Yunni Xia
Summary: Hierarchical clustering method HCNN effectively groups similar data by utilizing structural similarities in the nearest neighbor graph, identifying clusters and outliers while reducing the influence of obscure boundaries. The method merges clusters more efficiently by considering equivalence relations based on maximum similarity, leading to improved clustering efficiency and accuracy.
KNOWLEDGE-BASED SYSTEMS
(2021)
Article
Computer Science, Artificial Intelligence
Wang Zhou, Amin Ul Haq, Laixiang Qiu, Jehan Akbar
Summary: This article proposes a novel multi-view social recommendation scenario named MsRec, which improves recommendation performance and user experience by leveraging intricate inner relationships within social networks and various information sources. Experimental results demonstrate the significant advantages of MsRec over other benchmark recommender algorithms.
EXPERT SYSTEMS WITH APPLICATIONS
(2024)
Article
Statistics & Probability
Steve Hanneke, Aryeh Kontorovich, Sivan Sabato, Roi Weiss
Summary: The study expands a Bayesian-consistent learning algorithm in metric spaces and proves its universality. By defining essential separability, it shows the existence of a universally Bayesian-consistent learner, providing the first impossibility result for Bayesian consistency. The results completely characterize strong and weak universal Bayes consistency in metric spaces.
ANNALS OF STATISTICS
(2021)
Article
Statistics & Probability
Zhengliang Lu, Shengnan Ding, Fei Zhang, Rui Wang, Xuejun Wang
Summary: This work mainly investigates the consistency and strong convergence rate for the nearest neighbor density estimator based on -mixing random samples. The weak consistency, complete consistency, the rates of complete consistency, and strong consistency for the nearest neighbor estimator of density function based on -mixing random samples are established. The results obtained in the article extend some corresponding ones for independent samples.
COMMUNICATIONS IN STATISTICS-THEORY AND METHODS
(2022)
Article
Computer Science, Information Systems
Taushif Anwar, V Uma, Md Imran Hussain, Muralidhar Pantula
Summary: This paper analyzes the use of collaborative filtering to overcome the cold start and data sparsity issues in recommender systems. By generating user-item similarity matrix and prediction matrix, the CF approaches successfully address these problems and provide more relevant recommendations.
MULTIMEDIA TOOLS AND APPLICATIONS
(2022)
Article
Computer Science, Information Systems
Shitao Xiao, Yingxia Shao, Yawen Li, Hongzhi Yin, Yanyan Shen, Bin Cui
Summary: The paper introduces a novel collaborative filtering framework, LECF, which models interactions between users and items as edges and captures complex relationships. LECF predicts the existence probability of edges based on weighted similarities in a line graph and utilizes an efficient propagation algorithm for training and inference. Experimental results demonstrate that LECF outperforms state-of-the-art methods on four public datasets.
SCIENCE CHINA-INFORMATION SCIENCES
(2022)
Article
Mathematics
Lili Wang, Sunit Mistry, Abdulkadir Abdulahi Hasan, Abdiaziz Omar Hassan, Yousuf Islam, Frimpong Atta Junior Osei
Summary: The study presents an architecture for a recommendation system that transforms user items into narrow categories. It focuses on identifying movies that a user will likely watch based on their favorite items. The system prioritizes the shortest connections between item correlations and utilizes various methods to reduce data sparsity. It also demonstrates the ability to provide moderate recommendations from diverse perspectives.
Article
Mathematics
Xin Liu, Yi Wu, Wei Wang, Yong Zhu
Summary: In this paper, the authors establish a Bernstein inequality for m-asymptotically almost negatively associated random variables and further investigate the consistency of the nearest neighbor estimator of the density function. The results extend some existing findings in the literature. Numerical simulations are provided to validate the results.
Proceedings Paper
Computer Science, Artificial Intelligence
Kazi Omar Faruk, Anika Rahman, Sanjida Ali Shusmita, Md Sifat Ibn Awlad, Prasenjit Das, Md Humaion Kabir Mehedi, Shadab Iqbal, Annajiat Alim Rasel
Summary: With the increase in digitization, the number of internet users is growing along with the emergence of various internet-based businesses, one of which is modern ecommerce. The recommendation system plays a crucial role in the digital space and ecommerce by suggesting content to users based on their preferences. Collaborative filtering, such as the k nearest neighbor algorithm, is used to recommend items by finding similar users' preferences. In this study, the KNN collaborative filtering algorithm is applied to the Amazon Kindle Store Book review dataset, and a modified version considering expert users is proposed for more precise recommendations. The models are evaluated using RMSE, MAE, hit rate, and coverage, achieving outstanding results compared to the baseline algorithm.
19TH INTERNATIONAL SYMPOSIUM ON DISTRIBUTED COMPUTING AND ARTIFICIAL INTELLIGENCE
(2023)
Article
Computer Science, Artificial Intelligence
Chunsheng Cui, Meng Wei, Libin Che, Shouwen Wu, Erwei Wang
Summary: The continuous development of e-commerce recommendation systems allows people to obtain the products they need more effectively, saving both money and time. However, user reviews can sometimes be unclear, causing uncertainty in the information they provide. This paper introduces the concept of probabilistic linguistic term set (PLTS) as a statistical tool to depict the information in user reviews. Based on this, a hotel recommendation algorithm is proposed and tested using a case study.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)
Article
Computer Science, Cybernetics
Gourav Jain, Tripti Mahara, S. C. Sharma, Arun Kumar Sangaiah
Summary: Advancements in technology and internet penetration have led to the rise of online businesses. However, users often face information overload when making online purchases. This article proposes a new similarity measure and a cognitive similarity measure to improve the performance of recommendation systems. Experimental results demonstrate that these methods outperform existing techniques.
IEEE TRANSACTIONS ON COMPUTATIONAL SOCIAL SYSTEMS
(2022)
Article
Engineering, Marine
Minglong Zhang, Liang Huang, Yuanqiao Wen, Jinfen Zhang, Yamin Huang, Man Zhu
Summary: Ship location prediction has become popular in maritime transportation engineering for its benefits in safety supervision and security. Existing methods based on motion characteristics have uncertainties and cannot guarantee accuracy of trajectory predictions. This paper proposes an improved method using k-nearest neighbor (KNN) to predict ship locations. Experiments show that referencing multiple nearest neighbors with similar movements improves accuracy.
JOURNAL OF MARINE SCIENCE AND ENGINEERING
(2022)
Article
Computer Science, Artificial Intelligence
Daisy Monika Lal, Krishna Pratap Singh, Uma Shanker Tiwary
Summary: The proposed ICE and ICE-T metrics effectively measure information-coverage in automatically generated summaries. By utilizing pre-trained word embeddings, Part-Of-Speech based keyword extraction, and other techniques, they provide more accurate assessments compared to Rouge and BLEU.
EXPERT SYSTEMS WITH APPLICATIONS
(2022)