Project1_Regression

Fuzzy_Systems_Regression

TSK models that are using the hybrid method for training. For the task 1 use Airfoil Self-Noise Data Set. For the task 2 use Superconductivty Data Set

Task 1

Create 4 different TSK models (Models 1-4) that are using the hybrid method for training (Backpropagation and Least Squares Method)

• Use Airfoil Self-Noise Data Set.
• Use grid partition
• For Models 1-2 ---> Change the output function to constant and use 2 membership functions (Model 1) or 3 membership functions (Model 2)
• For Models 3-4 ---> Change the output function to linear and use 2 membership functions (Model 3) or 3 membership functions (Model 4)
• Use gbellmf as membership function type
• Train the TSK model with hybrid method (using Backpropagation and Least Squares Method)
• Evaluate the model
  • R2
  • RMSE
  • NMSE
  • NDEI

Task 2

• Use Superconductivty Data Set
• Use Subtractive Clustering (check many values of radius)
• Use Relieff algorithm to choose the most suitable features for training
• Compare metrics between models to find the best parameters with the help of grid search technique (for the number of features and the radius used in Subtractive Clustering