Xiao-Jun Li

Dr. Xiao-Jun Li

orcid

Research Focus & Metrics

Academic Career Length

10Years (2017-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
16 August 2018 in Journal of Medical Internet Research
Verified
Journal Article
Open access
JLJia Li
Jia Li
MLMinghui Liu
Minghui Liu
XLXiao-Jun Li
Xiao-Jun Li
LXLiu Xuan
Liu Xuan
JLJingfang Liu
Jingfang Liu
Journal of Medical Internet Research - International Scientific Journal for Medical Research, Information and Communication on the Internet #Preprint #PeerReviewMe: Warning: This is a unreviewed preprint. Readers are warned that the document has not been peer-reviewed by expert/patient reviewers or an academic editor, may contain misleading claims, and is likely to undergo changes before final publication, if accepted, or may have been rejected/withdrawn. Readers with interest and expertise are encouraged to sign up as peer-reviewer, if the paper is within an open peer-review period. Please cite this preprint only for review purposes or for grant applications and CVs (if you are the author). Background: Online physician reviews are invaluable gold mines that merit further investigation. Although many studies have explored the text information of physician reviews, very few have focused on developing a systematic topic taxonomy embedded in the physician reviews. The first step toward mining physician reviews is to determine how the natural structure or dimensions are embedded in reviews. Therefore, it is relevant to develop the topic taxonomy in a rigorous and systematic manner. Objective: In this study, we aim to (1) develop a hierarchical topic taxonomy to uncover the latent structure of physician reviews and (2) illustrate its application for mining patients’ interests based on the proposed taxonomy and algorithm. Methods: Data consisted of 122,716 physician reviews, including reviews of 8,501 doctors from a leading physician review website in China (haodf.com), which were collected between 2007 and 2015. Mixed methods, including a literature review, data-driven-based topic discovery, and human annotation were used to develop the physician review topic taxonomy. Results: The identified taxonomy includes three domains/high level categories and nine subtopics/low level categories. The physician-related domain includes the categories of medical ethics, medical competence, communication skills, medical advice and prescriptions. The patient-related domain includes the categories of the patient profile, symptoms, diagnosis and pathogenesis. The system-related domain includes the categories of financing and operation process. The F-measure of the proposed classification algorithm reaches 0.816 on average. Symptoms (Cohen’s d = 1.58, ∆u=0.216, t = 229.75, P = .000 < .001) is more often mentioned by patients with acute diseases, while communication skills (Cohen’s d = -0.29, ∆u=-0.038, t = -42.01, P = .000 < .001), financing (Cohen’s d = -0.68, ∆u=-0.098, t = -99.26, P = .000 < .001), and diagnosis and pathogenesis (Cohen’s d = -0.55, ∆u=-0.078, t = -80.09, P = .000 < .001) are more often mentioned by patients with chronic diseases. Patients with mild diseases were more interested in medical ethics (Cohen’s d = 0.25, ∆u=0.039, t = 8.33, P = .000 < .001), operation process (Cohen’s d = 0.57, ∆u=0.060, t = 18.75, P = .000 < .001), patient profile (Cohen’s d = 1.19, ∆u=0.132, t = 39.33, P = .000 < .001), and symptoms (Cohen’s d = 1.91, ∆u=0.274, t = 62.82, P = .000 < .001). Meanwhile, patients with serious diseases were more interested in medical competence (Cohen’s d = -0.99, ∆u=-0.165, t = -32.58, P = 0.0 < 0.001), medical advice and prescription (Cohen’s d = -0.65, ∆u=-0.082, t = -21.45, P = 0.0 < 0.001), financing (Cohen’s d = -0.26, ∆u=-0.018, t = -8.45, P = 0.0 < 0.001), and diagnosis and pathogenesis (Cohen’s d = -1.55, ∆u=-0.229, t = -50.93, P = 0.0 < 0.001). Conclusions: This mixed method approach that integrates literature reviews, data-driven topic discovery and human annotation is an effective and rigorous way to develop a physician review topic taxonomy. The proposed algorithm based on Labeled LDA (Latent Dirichlet Allocation) can achieve impressive classification results for mining patients’ interests. The mining results reveal significant differences for patients’ interests across different disease types, social economic development levels and hospital levels. Clinical Trial: N/A

18 May 2025 in Scientific Reports
Verified
Journal Article
Open access
PDF file
HSHaifeng Sun
Haifeng Sun
JYJunping Yao
Junping Yao
XLXiao-Jun Li
Xiao-Jun Li
YLYanfei Liu
Yanfei Liu
HGHongyang Gu
Hongyang Gu
Federated learning is widely used for collaborative training of human activity recognition models across multiple devices with limited local data. However, label noise caused by human and time constraints during data annotation is common and severely limits model performance. Existing studies mainly address this through client selection and sample filtering, but still face key limitations: (1) insufficient granularity in client quality evaluation; (2) aggregation methods ignoring data quality differences; (3) client drift under non-IID data distribution. To overcome these challenges of complex label noise and feature drift, this paper proposes LN-FHAR, a two-stage federated learning framework with label noise robustness. This framework effectively mitigates the coupling problem of noise and data heterogeneity by assessing client data quality and designing differentiated training strategies. In the client selection stage, clients are graded based on class-level loss analysis and a Gaussian Mixture Model. In the noise-robust training stage, reliable neighbors are introduced to collaboratively filter clean samples, and prototype regularization is employed to constrain the consistency between local models and global feature representations. Additionally, a data-aware aggregation method is designed, which assigns weights based on both the quality and quantity of client data, reducing the negative impact of noisy clients. Experimental results demonstrate that LN-FHAR has robustness and generalization ability in complex noise environments.

23 July 2024 in Knowledge-Based Systems
Verified
Journal Article
Closed access
XLXiao-Jun Li
Xiao-Jun Li
YSYi Su
Yi Su
JYJunping Yao
Junping Yao
HGHongyang Gu
Hongyang Gu
YJYibo Jiao
Yibo Jiao

1 January 2018 in Journal of Industrial and Management Optimization
Verified
Journal Article
Open access
PDF file
BCBinbin Cao
Binbin Cao
ZXZhongdong Xiao
Zhongdong Xiao
XLXiao-Jun Li
Xiao-Jun Li

1 June 2023 in IEEE Transactions on Geoscience and Remote Sensing
Verified
Journal Article
Closed access
YSYi Su
Yi Su
XLXiao-Jun Li
Xiao-Jun Li
JYJunping Yao
Junping Yao
CDChengrong Dong
Chengrong Dong
YWYao Wang
Yao Wang
Hyperspectral image classification has two key characteristics: spatial dependency and class imbalance. The synthetic oversampling methods for solving the between-class imbalance has the problem of spatial information loss to some extent. However, the feature space optimization methods for solving data within-class complex distribution problems have not been studied extensively. To remedy such issue, we propose in this paper an improved spectral-spatial feature rotation based SVM ensemble method (SSFRoF). More precisely, in the data preprocessing stage and from the perspective of sample space optimization, we devise a spectral-spatial neighborhood based adaptive synthetic sampling algorithm (SSAS) that focuses adaptively on hard-to-learn samples, with artificial samples interpolated in the spectral-spatial neighborhood to further utilize the spatial information. In feature rotation stage and from the perspective of feature space optimization, we innovatively discover that the class overlapping problem can be handled by combining feature extraction, so as to achieve the goal of reducing the difficulty of imbalanced classification. Besides, the SVM and voting procedure are used in the ensembled classification stage. We then conduct a series of experiments to compare the proposed SSFRoS with several baselines in terms of seven metrics. The results show that SSFRoS has significant advantages in dealing with imbalanced hyperspectral image classification problem. In addition, a detailed ablation study shows that the lowest class recall can be improved by 46.67% and 5.26% on Indian Pines and Pavia University respectively, which demonstrates that using SSAS can enhance the recognition of hard-to-learn classes.

20 July 2022 in Scientific Reports
Verified
Journal Article
Open access
PDF file
SLShuaitong Liu
Shuaitong Liu
XLXiao-Jun Li
Xiao-Jun Li
CHChanghua Hu
Changhua Hu
JYJunping Yao
Junping Yao
XHXiaoxia Han
Xiaoxia Han
JWJie Wang
Jie Wang
Spammer detection is essentially a process of judging the authenticity of users, and thus can be regarded as a classification problem. In order to improve the classification performance, multi-classifier information fusion is usually used to realize the automatic detection of spammers by utilizing the information from multiple classifiers. However, the existing fusion strategies do not reasonably take the uncertainty from the results of different classifiers (views) into account, and the relative importance and reliability of each classifier are not strictly distinguished. Therefore, in order to detect spammers effectively, this paper develops a novel multi-classifier information fusion model based on the evidential reasoning (ER) rule. Firstly, according to the user's characterization strategy, the base classifiers are constructed through the profile-based, content-based and behavior-based. Then, the idea of multi-classifier fusion is combined with the ER rule, and the results of base classifiers are aggregated by considering their weights and reliabilities. Extensive experimental results on the real-world dataset verify the effectiveness of the proposed model.