Computing and Information Systems - Research Publications
Permanent URI for this collection
Search Results
1 - 4 of 4
-
ItemDusting for fingerprints: Revealing patterns of online students’ behaviour( 2023-09-01)Hybrid learning strategies combine face-to-face instruction with online components. These hybrid environments rely heavily on online Learning Management Systems (LMSs) that serve as central hubs for learning materials. Depending on the adopted instructional strategy, students may be expected to complete certain tasks in the LMS. For instance, when adopting a flipped classroom strategy, students must watch videos, read material or complete quizzes prior the classroom time. Then, during classroom time, students focus on performing hands-on exercises. Students consistently engaging in the adopted strategy is key, as failing to do so can decrease the effectiveness of the adopted strategy. This paper presents an approach to monitor the changes of students’ behaviour over time. Using variants analysis, the method analyses data captured in an LMS and finds significant differences between time windows (e.g., two academic weeks). This can inform educators about changes in students’ engagement in the instructional strategy (e.g., students not completing some tasks). To showcase our method, we analyse a semester’s worth of data for a master’s subject implementing flipped classroom strategy.
-
ItemStrengthening Australia’s cybersecurity regulations and incentives: Response to the Department of Home Affairs Discussion Paper( 2021)The development of the regulatory and incentives framework is a key opportunity to align Australian enterprises’ cybersecurity practice with latest research, particularly on consumer protections, and emerging cyber threats and security challenges. The Australian Government has an essential role in establishing incentives to encourage best practice and consequences to combat poor practice. It will be increasingly important for government at all levels to act as a role model, by following best practice in the conduct of its public business.
-
Item
-
ItemEncoder-Decoder Generative Adversarial Nets for Suffix Generation and Remaining Time Predicationof Business Process Models( 2020)Predictive process monitoring aims to predict future characteristics of an ongoing process case, such as case outcome or remaining time till completion. Several deep learning models have been proposed to address suffix generation and remaining time prediction for ongoing process cases. Though they generally increase the prediction accuracy compared to traditional machine learning models, they still suffer from critical issues. For example, suffixes are generated by training a model on iteratively predicting the next activity. As such, prediction errors are propagated from one prediction step to the next, resulting in poor reliability, i.e., the ground truth and the generated suffixes may easily become dissimilar. Also, conventional training of neural networks via maximum likelihood estimation is prone to overfitting and prevents the model from generating sequences of variable length and with different activity labels. This is an unrealistic simplification as business process cases are often of variable length in reality. To address these shortcomings, this paper proposes an encoder-decoder architecture grounded on Generative Adversarial Networks (GANs), that generates a sequence of activities and their timestamps in an end-to-end way. GANs work well with differentiable data such as images. However, a suffix is a sequence of categorical items. To this end, we use the Gumbel-Softmax distribution to get a differentiable continuous approximation. The training works by putting one neural network against the other in a two-player game (hence the “adversarial” nature), which leads to generating suffixes close to the ground truth. From the experimental evaluation it emerges that the approach is superior to the baselines in terms of the accuracy of the predicted suffixes and corresponding remaining times, despite using a naive feature encoding and only engineering features based on control flow and events completion time.