| Abstract |
A smart sustainable city is an innovative city that use a smart technology and smart education besides other ecosystem's drivers in order to improve the quality of life. Several smart applications are developed by using different artificial intelligence models to enhance the education. A chatbot is considered as one of these advanced technologies that can be deployed in schools to enhance the education and learning processes in smart cities. However, high schools are the most crucial stage in student's life as they have higher stress levels regarding their future compared to other students. They need advices more than others do, for example, they need help to find the best-fit universities and courses that can fit with their interests and goals, however, most previous studies focused on providing chatbots for supporting the matriculated and new incoming students at universities. Therefore, to fill this gap, this study aims to develop a novel chatbot for supporting students in high schools particularly by using the Multinomial Naive-Bayes and Random Forest algorithms, which can be, able to understand the natural language of students' enquiries and accordingly, the intention class for each question will be predicted. The data in this study is comprised of 505 questions, which were obtained from multiple resources such as students' blogs, schools' counsellors, students, parents as well as from schools & universities' websites. The results in this study shows that the performance of the chatbot is improved by using the techniques of pre-processing and the feature extractions such as CountVectorizers and TF-IDF. Moreover, the results show that the Random Forest classifier performed better than Multinomial Naive-Bayes in all metrics. The performance of the models is checked by using different metrics such as accuracy, precision, recall and F1-score and all show high scores with exceeding 90%. However, the accuracy of Multinomial Naive-Bayes classifier achieved higher score when using CountVectorizers compared to using the TF-IDF. In the future work, this result will be investigated by re-evaluated the model with using a large corpus with students' enquiries. © 2023 University of Split, FESB. |
