ELREFAEI, I. Time Series Forecasting Using Machine Learning [online]. Brno: Vysoké učení technické v Brně. Fakulta elektrotechniky a komunikačních technologií. 2024.
The thesis, titled "Time Series Forecasting Using Machine Learning" focuses on implementing and applying machine learning (ML) techniques to datasets to predict future trends. However, the thesis has several shortcomings. It was submitted at the last moment before the deadline, which prevented thorough iterations with the supervisors. Moreover, it contains numerous formatting and language/grammar issues. Furthermore, the structure of the thesis is unclear, making it difficult for the reader to understand, especially from the "Table of Contents," which lacks crucial elements such as the main section numbers. Additionally, there is a disconnect between the theoretical and practical parts of the thesis. Despite initial agreements with the supervisors, the theoretical section on the ML models does not align well with the practical section, potentially confusing the reader. The references are weak and do not seem to be sourced from academic papers or books. The decision to share the code via Google Colab rather than GitHub was also not considered professional. However, the code presented in Colab indicates that the student's contribution to the theoretical aspects of the thesis is much more significant than the practical implementation. The student also described a simple application developed on Streamlit but failed to provide detailed information on its operation. Overall, the thesis does not meet the minimum quality criteria or the originally set objectives. Therefore, it requires significant revisions and ongoing consultation with the supervisor.
The presented thesis focuses on the application of machine learning in the area of time series prediction. The thesis starts with a pretty elaborated overview of the available algorithms, libraries, etc. It follows with data handling, training, and validation. The number of sources that the student used for this purpose adequate. The thesis then applies several of these algorithms on a testing dataset and evaluates the accuracy of the predictions. This process is documented with the code snippets and in my opinion it also reaches acceptable level of complexity. What I find as negative in the thesis are the following: a) the presentation (formal) side of the thesis does not follow the official guidelines and templates and the thesis comprise multiple formal mistakes; and b) the testing application that the student developed is super simple and it is barely mentioned in the thesis, this is where the student could have sold the thesis much better. In general, the student did achieve the goals of the thesis and given the above mentioned positives and negatives, I suggest the thesis to be given 70 points, i.e. grade C.
eVSKP id 153602