Research
Fast Control Gradient Algorithm for Simple and Multiple Linear Regression Model
Information is steadily increasing and hungry for knowledge. As the data grows, the world moves in on hunting knowledge with the help of analytics in the big data era. Flood data arising from diverse fields are described for automated learning technique of data analysis is intended as a machine learning, like classiffication and regression, which is a statistical method of predictive analysis. We proposed in this paper, gradient method with control step and Nestrove step called fast control gradient (FCG) algorithm for multiple linear regression (MLR), the quadratic convergence rate o(k2) of FCG algorithm are proved. FCG algorithm are applicate to a real dataset of wine quality for simple linear regression and dataset of combined cycle power plan (CCPP) for multiple linear regression. The numerical experiment, show that our approach FCG algorithm is faster than gradient descent (GD) algorithm.
