| This thesis describes the procedures used to identify the natural clusters formed by the students from Calculus that participated in the Quiz Project of the Mathematics Department at the University of Puerto Rico Mayaguez Campus.; The principal techniques for clustering and cluster validation are also discussed. In total four, clustering techniques with validation measures were employed.; The project was developed in stages. In the first stage data was collected from the Quiz system. For this purpose, an application was developed that inputs files containing quiz results, questionnaires and class results and outputs the information to matrix in which each row represented a student and each column an attribute. During the second stage the information was processed to eliminate unwanted data and weights were assigned to each datum associates with a student. A metric was employed that would help determine the similarities and dissimilarities between students.; To formulate an effective methodology to ascertain the clusters that were present in the actual data, datasets consisting of a number of n-dimensional normal distribution with a variety of means and standard deviations were prepared to simulate real data. The simulated datasets were grouped with the cluster algorithms and validation measures were used to determinate the quality of the grouping. The methodology developed with the simulated data was then applied to find natural cluster existing in the real data. |
