Estimation of the number of cluster in nonnegative matrix factorizations

Using different methods and techniques to achieve the best results and create synergies between them can be helpful in many issues. Educational data mining is one of the relatively new fields that can be used to solve educational problems, especially problems in the field of measurement. But before using these methods, it should be as familiar as possible and its problems, disadvantages and advantages should be examined. In this study, the aim is to investigate the technique of nonnegative matrix factorizations and how to determine the number of clusters before implementing the model. The research method is descriptive and the study population is all those present at the entrance exam for mathematics and technical sciences in 1398, of which 5,000 people were randomly selected by the country's assessment and education organization and provided to the researcher. The research tool is math questions and entrance exam geometry. The results of this analysis showed that there is a difference in estimating the number of clusters of math questions, but in the case of geometry questions, the results of all methods were the same. Due to the observed differences, it is suggested that in future research with the help of data simulation, this issue will be examined in more detail.