Abstract:

The introduction of cutting-edge digital tools and the development of new technologies for blended learning approaches have added a new dimension to the field of distance education. Several educational institutions have developed and adopted hybrid learning programs, particularly in the wake of recent pandemic. Various virtual environments are utilized for video chatting, file sharing, assignment submission, and screen sharing, thereby providing students and instructors with a shared virtual platform. In the pandemic era, the objects used in traditional architecture studio education, such as sketches, physical models, and plans, can only be examined through the use of digital monitors and tools. The transformation of architecture studio education, which is situated at the intersection of theoretical and practical knowledge, into a fully blended format is an experiment for both students and instructors. At this time, it is crucial to investigate the issues and logical solutions encountered during this process, which has been ongoing for approximately a year and may continue in the future. Artificial neural networks (ANNs) are highly effective at finding and labeling patterns in large, complicated datasets. Data mining methods like artificial neural nets are also put to use in the assessment of student performance and the forecasting of instructional practices. Similar methods and models might be applied here to further examine educational records. Generalized Regression Neural Networks (GRNNs) and Multilayer Feed-Forward Neural Nets (MLFNNs) were trained in this study to measure the academic performance of students using a dataset containing numeric data. The dataset consists of phases of training (80%), and testing (20%). The input parameters include the working grade of the group, the level of detail of the final 3D digital model, performance and active participation, open-semi-open and closed space relationships, and cardboard model fabrication styles. An output parameter is the final grade. In the results section, Multilayer Feed-Forward Neural Nets (MLFNNs) with different number of nodes (2-3-4-5-6) and Generalized Regression Neural Networks (GRNNs) are compared. According to the results, the performance of the GRNN model is superior. Root mean square error (RMSE), standard deviation of absolute error, bad prediction rate, mean absolute error (MAE), and R-squared values are obtained after choosing the best network. In addition, a sensitivity analysis and variable impact analysis were performed. Future experimental processes in design education may benefit academically from the use of blended learning methodologies to studio culture and activities and the evaluation of their academic consequences. Educational data mining technologies can help students better comprehend and make sense of the interconnections between ideas, as well as comprehend and anticipate the results of procedures.

Keywords:

Blended Learning Model, Online Design Education, Artificial Neural Network.