Evidence Combination for Pattern Classification With Missing Data DOI Creative Commons
Liu Lizhuo

Research Square (Research Square), Journal Year: 2023, Volume and Issue: unknown

Published: Aug. 11, 2023

Abstract It remains a challenging problem to classify the incomplete patterns with randomly missing values. In some applications, it is difficult for us collect complete attributes of target due complex sensing environment, and observed are all more or less order well such patterns, we propose new classification method based on evidence combination. Because considered attribute values, hard accurately estimate values object classify. We using each available respectively. The K-nearest neighbors (K-NN) found in training data space according attribute, K Basic belief assignments (BBA) constructed corresponding K-NN. BBA reflects degree belonging class. Then two step combination strategy developed combining BBA's. BBA's associated same class combined by classical DS rule at first. When K-NN belong different classes, previous results classes may highly conflict, they further PCR5 rule, which can manage high conflict via proper conflicting masses redistribution. other one similar way. multiple collected, weighted employed fuse these results. weighting factors optimized minimizing an error criterion. finally classified depending this result. Experimental various sets show effectiveness proposed comparing related methods.

Language: Английский

mdatagen: A python library for the artificial generation of missing data DOI
Arthur Dantas Mangussi, Miriam Seoane Santos, Filipe Loyola Lopes

et al.

Neurocomputing, Journal Year: 2025, Volume and Issue: unknown, P. 129478 - 129478

Published: Jan. 1, 2025

Language: Английский

Citations

1
Interpretable knowledge-guided framework for modeling reservoir water-sensitivity damage based on Light Gradient Boosting Machine using Bayesian optimization and hybrid feature mining DOI
Keming Sheng, Guancheng Jiang, Mingliang Du

et al.

Engineering Applications of Artificial Intelligence, Journal Year: 2024, Volume and Issue: 133, P. 108511 - 108511

Published: April 27, 2024

Language: Английский

Citations

8
Analysis of the summer thermal comfort indices in İstanbul DOI Creative Commons
Merve Yılmaz, Yiğitalp Kara, Hüseyi̇n Toros

et al.

International Journal of Biometeorology, Journal Year: 2024, Volume and Issue: 68(7), P. 1327 - 1342

Published: April 24, 2024

Thermal indices and thermal comfort maps have great importance in developing health-minded climate action strategies livable urban layouts. Especially cities where vulnerability to heatwaves is high, it necessary detect the most appropriate indicators for regional characteristics planning with respect comfort. The aim of study examine as by relating meteorological parameters spatial features. Atmospheric variables including air temperature, wind speed, cloud cover, relative humidity data were obtained from 30 stations located districts having different climatic Heat stress levels apparent temperature (AT), heat index (HI), wet bulb globe (WBGT), physiological equivalent (PET), universal (UTCI), perceived (PT) calculated associated parameters. been created daily mean maximum values all indices. As a result, strongest correlation are (T

Language: Английский

Citations

4
Tracking-removed neural network with graph information for classification of incomplete data DOI Creative Commons
Xiaochen Lai, Zheng Zhang, Hui Chen

et al.

Applied Intelligence, Journal Year: 2025, Volume and Issue: 55(4)

Published: Jan. 2, 2025

Language: Английский

Citations

0
Multilevel Multigroup Structural Equation Modeling In A Single-Level Framework DOI Creative Commons
Julia-Kim Walther, Martin Hecht, Benjamin Nagengast

et al.

Structural Equation Modeling A Multidisciplinary Journal, Journal Year: 2025, Volume and Issue: unknown, P. 1 - 32

Published: Jan. 6, 2025

Heterogeneity of variance is more than a statistical nuisance when parameters are substantial interest. In multilevel modeling (e.g. students within classes), for instance, the inclusion discrete variables at between-cluster level school type) may lead to detection differences between variances within-cluster students' performance in test). The resulting heterogeneous lower high schools compared grammar schools) have potential inform research and practice on educational effectiveness). Along lines 'people too', we demonstrate how single-level formulation structural equation models, wide format approach (Barendse & Rosseel, Citation2020; Mehta Neale, Citation2005), can be used combination with multigroup order obtain estimates. We provide evidence proposed WFmultigroup approaches' accuracy by means simulation study showcase its application an empirical illustration lavaan package R.

Language: Английский

Citations

0
Impute-VSS: A comprehensive web-based visualization and simulation suite for comparative data imputation and statistical evaluation DOI
Vivek Shrivastava, Shekhar Shukla

SoftwareX, Journal Year: 2025, Volume and Issue: 30, P. 102130 - 102130

Published: March 15, 2025

Language: Английский

Citations

0
Multimodal data imputation and fusion for trustworthy fault diagnosis of mechanical systems DOI
Jie Zhang, Yun Kong, Qinkai Han

et al.

Engineering Applications of Artificial Intelligence, Journal Year: 2025, Volume and Issue: 150, P. 110663 - 110663

Published: March 28, 2025

Language: Английский

Citations

0
Siamese Autoencoder Architecture for the Imputation of Data Missing Not at Random DOI Creative Commons
Ricardo Cardoso Pereira, Pedro Henriques Abreu, Pedro Pereira Rodrigues

et al.

Journal of Computational Science, Journal Year: 2024, Volume and Issue: 78, P. 102269 - 102269

Published: March 24, 2024

Missing data is an issue that can negatively impact any task performed with the available and it often found in real-world domains such as healthcare. One of most common strategies to address this perform imputation, where missing values are replaced by estimates. Several approaches based on statistics machine learning techniques have been proposed for purpose, including deep architectures generative adversarial networks autoencoders. In work, we propose a novel siamese neural network suitable which call Siamese Autoencoder-based Approach Imputation (SAEI). Besides having autoencoder architecture, SAEI also has custom loss function triplet mining strategy tailored issue. The approach compared seven state-of-the-art imputation methods experimental setup comprises 14 heterogeneous datasets healthcare domain injected Not At Random at rate between 10% 60%. results show significantly outperforms all remaining experimented settings, achieving average improvement 35%. This work extension article Autoencoder-Based Data (Pereira, et al. 2023) presented International Conference Computational Science 2023. It includes new experiments focused runtime, generalization capabilities, classification tasks, method induces best results, improving F1 scores 50% used datasets.

Language: Английский

Citations

3
Feature-wise attention based boosting ensemble method for fraud detection DOI
Ruihao Cao, Junli Wang,

Mingze Mao

et al.

Engineering Applications of Artificial Intelligence, Journal Year: 2023, Volume and Issue: 126, P. 106975 - 106975

Published: Aug. 23, 2023

Language: Английский

Citations

7
Computing Random Forest-distances in the presence of missing data DOI Open Access
Manuele Bicego, Ferdinando Cicalese

ACM Transactions on Knowledge Discovery from Data, Journal Year: 2024, Volume and Issue: 18(7), P. 1 - 18

Published: April 8, 2024

In this article, we study the problem of computing Random Forest-distances in presence missing data. We present a general framework which avoids pre-imputation and uses an agnostic way information contained input points. centre our investigation on RatioRF, RF-based distance recently introduced context clustering shown to outperform most known measures. also show that same can be applied several other state-of-the-art measures provide their extensions data case. significant empirical evidence effectiveness proposed framework, showing extensive experiments with RatioRF 15 datasets. Finally, positively compare method many alternative literature distances, computed values.

Language: Английский

Citations

1