In this paper we apply supervised machine learning algorithms to
automatically classify the text of students' reflective learning journals from an introductory Java programming module with the aim of identifying students who need help with their understanding of the topic they are reflecting on. Such a system could alert teaching staff to students who may need an intervention to support their learning.
Several different classifier algorithms have been validated on the training data set to find the best model in two situations; with equal cost for a positive or negative classification and with cost sensitive classification. Methods were used to identify those individual parameters which maximise the performance of each algorithm. Precision, recall and F1-score, as well as confusion matrices were used to understand the behaviour of each classifier and choose the one with the best performance.
The classifiers that obtained the best results from the validation were then evaluated on a testing data set containing different data to that used for training.
We believe that although the results could be improved with further work, our initial results show that machine learning could be applied to students' reflective writing to assist staff in identifying those students who are struggling to understand the topic.
|Title of host publication||Machine Learning, Optimization, and Data Science - 5th International Conference, LOD 2019, Proceedings|
|Editors||Giuseppe Nicosia, Panos Pardalos, Renato Umeton, Giovanni Giuffrida, Vincenzo Sciacca|
|Number of pages||12|
|Publication status||Published - 3 Jan 2020|
|Event||5th International Conference on Machine Learning, Optimization, and Data Science, LOD 2019 - Siena, Italy|
Duration: 10 Sep 2019 → 13 Sep 2019
|Name||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Conference||5th International Conference on Machine Learning, Optimization, and Data Science, LOD 2019|
|Period||10/09/19 → 13/09/19|