An embedding approach for analyzing the evolution of research topics with a case study on computer science subdomains.

The study of topic evolution aims to analyze the behavior of different research fields by utilizing various features such as the relationships between articles. In recent years, many published papers consider more than one field of study which has led to a significant increase in the number of inter-field and interdisciplinary articles. Therefore, we can analyze the similarity/dissimilarity and convergence/divergence of research fields based on topic analysis of the published papers. Our research intends to create a methodology for studying the evolution of the research fields. In this paper, we propose an embedding approach for modeling each research topics as a multidimensional vector. Using this model, we measure the topic's distances over the years and investigate how topics evolve over time. The proposed similarity metric showed many advantages over other alternatives (such as Jaccard similarity) and it resulted in better stability and accuracy. As a case study, we applied the proposed method to subsets of computer science for experimental purposes, and the results were quite comprehensible and coherent.

Classification of complex networks based on similarity of topological network features.

Over the past few decades, networks have been widely used to model real-world phenomena. Real-world networks exhibit nontrivial topological characteristics and therefore, many network models are proposed in the literature for generating graphs that are similar to real networks. Network models reproduce nontrivial properties such as long-tail degree distributions or high clustering coefficients. In this context, we encounter the problem of selecting the network model that best fits a given real-world network. The need for a model selection method reveals the network classification problem, in which a target-network is classified into one of the candidate network models. In this paper, we propose a novel network classification method which is independent of the network size and employs an alignment-free metric of network comparison. The proposed method is based on supervised machine learning algorithms and utilizes the topological similarities of networks for the classification task. The experiments show that the proposed method outperforms state-of-the-art methods with respect to classification accuracy, time efficiency, and robustness to noise.

Distance metric learning for complex networks: towards size-independent comparison of network structures.

Real networks show nontrivial topological properties such as community structure and long-tail degree distribution. Moreover, many network analysis applications are based on topological comparison of complex networks. Classification and clustering of networks, model selection, and anomaly detection are just some applications of network comparison. In these applications, an effective similarity metric is needed which, given two complex networks of possibly different sizes, evaluates the amount of similarity between the structural features of the two networks. Traditional graph comparison approaches, such as isomorphism-based methods, are not only too time consuming but also inappropriate to compare networks with different sizes. In this paper, we propose an intelligent method based on the genetic algorithms for integrating, selecting, and weighting the network features in order to develop an effective similarity measure for complex networks. The proposed similarity metric outperforms state of the art methods with respect to different evaluation criteria.

Generative model selection using a scalable and size-independent complex network classifier.

Real networks exhibit nontrivial topological features, such as heavy-tailed degree distribution, high clustering, and small-worldness. Researchers have developed several generative models for synthesizing artificial networks that are structurally similar to real networks. An important research problem is to identify the generative model that best fits to a target network. In this paper, we investigate this problem and our goal is to select the model that is able to generate graphs similar to a given network instance. By the means of generating synthetic networks with seven outstanding generative models, we have utilized machine learning methods to develop a decision tree for model selection. Our proposed method, which is named "Generative Model Selection for Complex Networks," outperforms existing methods with respect to accuracy, scalability, and size-independence.