Period Estimation of an Almost Periodic Signal Using Persistent Homology With Application to Respiratory Rate Measurement

Erden, Fatih; Çetin, A. Enis

doi:10.1109/lsp.2017.2699924

Cited by 11 publications

(1 citation statement)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The basic idea is giving shape to data by replacing it with a family of simplicial complexes, which can be considered as a higher dimensional generalization of graphs, and then studying the topology of the resulted shapes to gain knowledge about it [5]. It has found applications in many different domains such as biological systems [6], computer vision [7], social network analysis [8] and signal processing [9]. Finally, we employ machine learning algorithms on these features to first reduce the dimension of the features and then classify the users as bot vs. human.…”

Section: Introductionmentioning

confidence: 99%

Bot Detection on Social Networks Using Persistent Homology

Nguyen

Aktaş

Akbaş

2020

MCA

View full text Add to dashboard Cite

The growth of social media in recent years has contributed to an ever-increasing network of user data in every aspect of life. This volume of generated data is becoming a vital asset for the growth of companies and organizations as a powerful tool to gain insights and make crucial decisions. However, data is not always reliable, since primarily, it can be manipulated and disseminated from unreliable sources. In the field of social network analysis, this problem can be tackled by implementing machine learning models that can learn to classify between humans and bots, which are mostly harmful computer programs exploited to shape public opinions and circulate false information on social media. In this paper, we propose a novel topological feature extraction method for bot detection on social networks. We first create weighted ego networks of each user. We then encode the higher-order topological features of ego networks using persistent homology. Finally, we use these extracted features to train a machine learning model and use that model to classify users as bot vs. human. Our experimental results suggest that using the higher-order topological features coming from persistent homology is promising in bot detection and more effective than using classical graph-theoretic structural features.

show abstract