Topic Cluster Identification Using Spectral Clustering in Academic Research Organization Platforms

To plan a successful academic research, it is necessary to find relevant groups of topics, which will facilitate better information search and discovery of knowledge on research platforms. Conventional clustering methods tend to have difficulties with high-dimensional, sparse, and complex relations between research articles, which translates to poor grouping and low interpretability. Current approaches, including hierarchical or k-means clustering, can be unsuccessful in encompassing non-linear structures as well as global connectivity in academic data, and thus produce incorrect or discontinuous groups of topics. This research proposes a model to deal with these issues; the model involves the Spectral Clustering Algorithm (SCA), which uses the eigenvalues of similarity matrices to naturally group data. SCA plays a crucial role in extracting detailed insights into complex relations among articles since it can reduce high-dimensional research data to a lower-dimensional spectral space, achieving coherence and more sense-making topic clusters. The suggested approach is implemented on the organization platforms of academic research to automatically cluster articles, authors, and research topics. This improves navigation, recommendation, and discovery of relevant research materials for scholars and institutions. Through experimental analysis, it has been shown that SCA is better at clustering patterns compared to the traditional clustering protocols in cluster cohesion, separation, and overall accuracy. The results show the enhancement of topic recognition, which helps to organize academic knowledge better and allows making research management decisions based on the results.

Ҳали таржима қилинмаган