Kristel Vignery

Doctorante et Assistante de cours en Gestion (USL-B)

Projet de Thèse :

"Integrating Knowledge Management In Prediction Techniques : Impact Of Online Social Networks On Academic Achievement"

Summary of project

The impact of individual and situational characteristics in the prediction of achievement in school and university has been studied through numerous researches. On the other hand, social networks and online tools are now integral part of our lives, and it becomes more and more necessary to include these features in the achievement's prediction. Two methodological aspects of the research conducted on such increasingly complex networks have drawn our attention. The first aspect covers knowledge management (e.g., through the use of ontology) that allows us to build and maintain a body of knowledge to which this information contributes. Second, data analysis (e.g., data mining tools, clustering techniques), which allows us to reduce large amounts of data to concise information.

The objective of this research project would be to cluster students present in social networks and online social networks (especially an official online platform dedicated to university classes) given their links with other students. To achieve this goal, we would look for the most appropriate clustering techniques to analyze a given set of data, among segmentation’s algorithms (e.g., Hierarchical agglomerative clustering, K-means) and different distances/similarities between the nodes of a graph (e.g., Euclidean Commute Time Distance, Minimax Path-Based Dissimilarity Measure). In order to increase the accuracy of our predictions, we would use the outcomes of these analyses to validate a knowledge base (i.e., an ontology) of the network, built in order to represent a mental and theoretical model which concerns a students community.

In a final stage, these clusters will be used to analyze and predict the achievement of students composing the network. Their success will also be studied according their individual characteristics, by means of modeling techniques (e.g., hierarchical modeling, logistic regression ...). At a later stage, we might be able to refine and apply these techniques to predict other behaviors. A potential outcome might be that, based on these research results, universities could develop politics promoting an optimal use of official online tools.


Comments are closed.