Dataset Information

Multilabel user classification using the community structure of online networks.

ABSTRACT: We study the problem of semi-supervised, multi-label user classification of networked data in the online social platform setting. We propose a framework that combines unsupervised community extraction and supervised, community-based feature weighting before training a classifier. We introduce Approximate Regularized Commute-Time Embedding (ARCTE), an algorithm that projects the users of a social graph onto a latent space, but instead of packing the global structure into a matrix of predefined rank, as many spectral and neural representation learning methods do, it extracts local communities for all users in the graph in order to learn a sparse embedding. To this end, we employ an improvement of personalized PageRank algorithms for searching locally in each user's graph structure. Then, we perform supervised community feature weighting in order to boost the importance of highly predictive communities. We assess our method performance on the problem of user classification by performing an extensive comparative study among various recent methods based on graph embeddings. The comparison shows that ARCTE significantly outperforms the competition in almost all cases, achieving up to 35% relative improvement compared to the second best competing method in terms of F1-score.

SUBMITTER: Rizos G

PROVIDER: S-EPMC5344401 | biostudies-literature | 2017

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Multilabel user classification using the community structure of online networks.

Rizos Georgios G Papadopoulos Symeon S Kompatsiaris Yiannis Y

PloS one 20170309 3

We study the problem of semi-supervised, multi-label user classification of networked data in the online social platform setting. We propose a framework that combines unsupervised community extraction and supervised, community-based feature weighting before training a classifier. We introduce Approximate Regularized Commute-Time Embedding (ARCTE), an algorithm that projects the users of a social graph onto a latent space, but instead of packing the global structure into a matrix of predefined ra ...[more]

PMID: 28278242

Similar Datasets

Project description:Standard reference terminology of diagnoses and risk factors is crucial for billing, epidemiological studies, and inter/intranational comparisons of diseases. The International Classification of Disease (ICD) is a standardized and widely used method, but the manual classification is an enormously time-consuming endeavor. Natural language processing together with machine learning allows automated structuring of diagnoses using ICD-10 codes, but the limited performance of machine learning models, the necessity of gigantic datasets, and poor reliability of terminal parts of these codes restricted clinical usability. We aimed to create a high performing pipeline for automated classification of reliable ICD-10 codes in the free medical text in cardiology. We focussed on frequently used and well-defined three- and four-digit ICD-10 codes that still have enough granularity to be clinically relevant such as atrial fibrillation (I48), acute myocardial infarction (I21), or dilated cardiomyopathy (I42.0). Our pipeline uses a deep neural network known as a Bidirectional Gated Recurrent Unit Neural Network and was trained and tested with 5548 discharge letters and validated in 5089 discharge and procedural letters. As in clinical practice discharge letters may be labeled with more than one code, we assessed the single- and multilabel performance of main diagnoses and cardiovascular risk factors. We investigated using both the entire body of text and only the summary paragraph, supplemented by age and sex. Given the privacy-sensitive information included in discharge letters, we added a de-identification step. The performance was high, with F1 scores of 0.76-0.99 for three-character and 0.87-0.98 for four-character ICD-10 codes, and was best when using complete discharge letters. Adding variables age/sex did not affect results. For model interpretability, word coefficients were provided and qualitative assessment of classification was manually performed. Because of its high performance, this pipeline can be useful to decrease the administrative burden of classifying discharge diagnoses and may serve as a scaffold for reimbursement and research applications.

Project description:Introduction:User generated content (UGC) is a valuable but underutilized source of information about individuals who participate in online cessation interventions. This study represents a first effort to passively detect smoking status among members of an online cessation program using UGC. Methods:Secondary data analysis was performed on data from 826 participants in a web-based smoking cessation randomized trial that included an online community. Domain experts from the online community reviewed each post and comment written by participants and attempted to infer the author's smoking status at the time it was written. Inferences from UGC were validated by comparison with self-reported 30-day point prevalence abstinence (PPA). Following validation, the impact of this method was evaluated across all individuals and time points in the study period. Results:Of the 826 participants in the analytic sample, 719 had written at least one post from which content inference was possible. Among participants for whom unambiguous smoking status was inferred during the 30 days preceding their 3-month follow-up survey, concordance with self-report was almost perfect (kappa = 0.94). Posts indicating abstinence tended to be written shortly after enrollment (median = 14 days). Conclusions:Passive inference of smoking status from UGC in online cessation communities is possible and highly reliable for smokers who actively produce content. These results lay the groundwork for further development of observational research tools and intervention innovations. Implications:A proof-of-concept methodology for inferring smoking status from user generated content in online cessation communities is presented and validated. Content inference of smoking status makes a key cessation variable available for use in observational designs. This method provides a powerful tool for researchers interested in online cessation interventions and establishes a foundation for larger scale application via machine learning.

Project description:BACKGROUND:According to the American Diabetes Association, there are approximately 30.3 million Americans with diabetes, and the incidence is growing by nearly 1.5 million cases per year. These individuals are at particularly high risk of developing secondary comorbid conditions related to diabetes and aging. Nearly 45% of individuals aged 65 to 75 years use social media, and this number is steadily growing. The use of social media provides the opportunity to assess the perceptions and needs of this population. OBJECTIVE:The purpose of this study was to examine stakeholder perceptions of successful aging with diabetes. METHODS:This study presents a retrospective analysis of a tweet chat focused on aging with diabetes. Tweets were collected using Symplur Signals data analytics software (Symplur LLC) and analyzed for content analysis, sentiment, and participant demographics. Two authors reviewed discussion posts for accuracy of analysis. RESULTS:A total of 59 individuals participated in this tweet chat generating 494 tweets and nearly 2 million impressions. Most (36/59, 63%) tweet chat participants were people living with diabetes; 25% (14/59) were caregivers and advocates. Seven countries were represented in the conversation. A majority (352/494, 71.3%) of the tweets indicated positive sentiment related to aging with diabetes. Five major themes emerged from the qualitative analysis: (1) personal decline now and in the future, (2) limited access to treatment, (3) inability to provide self-care, (4) health care provider capacity to support aging with diabetes, and (5) life-long online peer health support to facilitate diabetes management. CONCLUSIONS:Individuals with diabetes are living longer and want to be supported with specialized care and access to technology that will allow them to successfully age. Aging- and diabetes-related changes may complicate diabetes management into old age. People with diabetes desire options including aging in place; therefore, special training for care partners and health care providers who care for older adults is needed.

Dataset Information

Multilabel user classification using the community structure of online networks.

Publications

Multilabel user classification using the community structure of online networks.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure