Comparative analysis of innovative neural networks techniques for time series forecasting
External reference persons Alessandro Aliberti (firstname.lastname@example.org)
Thesis type EXPERIMENTAL
Description This thesis aims at carrying out a comparative analysis of the most innovative and used machine learning techniques for time series forecasting. For this purpose, the student will have to investigate the state of the art on neural network techniques and identify the most effective methodologies (e.g. ARX, RNN, LSTM, etc.). Starting from a common dataset provided by Internet-of-Things devices, these methodologies will be implemented on different platforms (e.g. Matlab, Tensorflow, PyTorch, etc.) in order to measure the different performances in terms of computational cost, execution time and prediction error.
Deadline 30/03/2019 PROPONI LA TUA CANDIDATURA