فرشاد رمضانخاني

The advancement of computer technology and sensor systems has facilitated the production and collection of diverse and large-scale data at high speeds, creating both opportunities for innovation and challenges for engineers. In regression models, the input and output data are not always numerical and encompass various types of information. This thesis presents a comprehensive study of regression models where the variables are represented as complex structured data, including tensor or functional data. The quality of the output process examined in this thesis is predicted with high accuracy using regression models developed based on adjustable and influential input variables that are not solely numerical. These models are employed to model nonlinear and complex relationships between variables and to identify hidden patterns within the data, as well as deep dependencies among variables, to enhance decision-making processes in various fields, particularly in machine learning, computer vision, complex signal analysis, and multidimensional data processing. The friction stir welding method is used as a case study in this thesis to validate the results. Statistical tests specific to each developed regression model have been designed for eva‎luation purposes, with simulations conducted using R software.

رگرسيون , داده تنسوري , داده تابعي , چندجويباره , جوشكاري اصطكاكي اغتشاشي

regression , Tensorial Data , functional data , Multi-Channel , friction stir welding

Farshad Ramezankhani

Dr. Rasool Noorosana