Using Fourier coefficients in time series analysis for student performance prediction in blended learning environments

Abstract

In this work, it is shown that student access time series generated from Moodle log files contain information sufficient for successful prediction of student final results in blended learning courses. It is also shown that if time series is transformed into frequency domain, using discrete Fourier transforms (DFT), the information contained in it will be preserved. Hence, resulting periodogram and its DFT coefficients can be used for generating student performance models with the algorithms commonly used for that purposes. The amount of data extracted from log files, especially for lengthy courses, can be huge. Nevertheless, by using DFT, drastic compression of data is possible. It is experimentally shown, by means of several commonly used modelling algorithms, that if in average all but 5–10% of most intensive and most frequently used DFT coefficients are removed from datasets, the modelling with the remained data will result with the increase of the model accuracy. Resulting accuracy of the calculated models is in accordance with results for student performance models calculated for different dataset types reported in literature. The advantage of this approach is its applicability because the data are automatically collected in Moodle logs.

Learning Analytics for Blended Learning: A Systematic Review of Theory, Methodology, and Ethical Considerations

“…[36,52,58,80]. Numerous studies explored relations between trace data, performance and SRL using self-reports [30]; some in combination with other theories, for example theories of motivation [20] socio-cultural perspectives [31], Self-Determination Theory and the Control-Value Theory of achievement emotions [86][87][88]. 5.…”

“…However, the portability of predictive models across courses remains low [23,32]. Since prediction is entirely dependent on the supplied data, studies identified that LMS variables (e.g., access time), engagement indicators, self-regulated learning (e.g., self-efficacy and test anxiety), and collaborative learning (e.g., social stability, and time spent on task) variables have reliable predictive power due to their positive correlation with students achievements [2,30,71,80,90,91,102]. Nevertheless, for some courses (e.g., graphic design) tracking data became useless because different patterns exist in the effect of individual data variables [32].…”

Data Mining for Student Outcome Prediction on Moodle: a systematic mapping

“…As EDM is still a new research domain, the great majority of the papers present data mining analysis of specific situations, experimenting with different types of methods [Hung et al 2016], [Zorrilla and Garcia-Saiz 2014], [Hu et al 2014], [Márquez-Vera et al 2013], [Romero et al 2013a], [Jovanovic et al 2012], [Kotsiantis 2012], [Minaei-Bidgoli et al 2003], [Thai-Nghe et al 2009], [Pardos et al 2012], [Romero et al 2008], [Sharma and Mavani 2011a] Neural Network [Gamulin et al 2016], [Cambruzzi et al 2015], [Shana and Abdulla 2015], [Sorour et al 2014], [Romero et al 2013a], [Kotsiantis 2012], [Sharma and Mavani 2011a], [Lykourentzou et al 2009b], [Lykourentzou et al 2009a] Bayesian Classification [Gamulin et al 2016], [Zorrilla and Garcia-Saiz 2014], [Romero et al 2013b], [Sharma and Mavani 2011b], [Sharma and Mavani 2011a], [Kotsiantis et al 2010], [Thai-Nghe et al 2009], [ Minaei-Bidgoli et al 2003] Support Vector Machines [Gamulin et al 2016], [Kotsiantis 2012], [Lykourentzou et al 2009b], [Thai-Nghe et al 2009] Genetic Algorithm [Márquez-Vera et al 2016], [Xing et al 2015], [Romero et al 2013a], [Zafra and Ventura 2012], [Zafra et al 2011], [Zafra and Ventura 2009],…”

“…Partitioning Methods [Gamulin et al 2016], [Moradi et al 2014], [Sorour et al 2014], [Romero et al 2013b], [Jovanovic et al 2012], [Mogus et al 2012], [López et al 2012], [Pardos et al 2012], [Kotsiantis et al 2010], [Obadi et al 2010], [Minaei-Bidgoli et al 2003] Association…”

“…The most widely used Decision Tree algorithm is C4.5 [Zorrilla and Garcia-Saiz 2014], [Romero et al 2013a], [Hu et al 2014], followed by Simple Cart [Zorrilla and Garcia-Saiz 2014], [Romero et al 2013a], [Hu et al 2014], and Random Trees [Romero et al 2008], [Romero et al 2013a], [Hu et al 2014 [Romero et al 2013a]. Other popular algorithms are: K-means clustering [Moradi et al 2014], [Jovanovic et al 2012], [Pardos et al 2012], [Mogus et al 2012], [Sorour et al 2014]; K-nearest neighbor (KNN) [Zorrilla and Garcia-Saiz 2014], [Kotsiantis et al 2010], [Minaei-Bidgoli et al 2003], [Gamulin et al 2016]; and JRip [Zorrilla and Garcia-Saiz 2014], [Márquez-Vera et al 2013].…”