Skip to content

TSK models using the hybrid method for classification problems

License

Notifications You must be signed in to change notification settings

ArmandoDomi/Fuzzy_Systems_Classification

Repository files navigation

Fuzzy_Systems_Classification

TSK models that are using the hybrid method for training. For the task 1 use Haberman's dataset. For the task 2 use Epileptic Seizure Recognition Data Set

Task 1 Model 1

  • Use Subtractive Clustering
  • Class Independent for clusterInfluenceRange = 0.7
  • Change the output function to constant
  • Train the TSK model with hybrid method (Backpropagation and Least Squares Method)
  • Evaluate the model
    • Error matrix
    • Producer’s accuracy – User’s accuracy
    • Overall accuracy
    • K

Task 1 Model 2

  • Use Subtractive Clustering
  • Class Independent for clusterInfluenceRange = 0.9
  • Change the output function to constant
  • Train the TSK model with hybrid method (Backpropagation and Least Squares Method)
  • Evaluate the model
    • Error matrix
    • Producer’s accuracy – User’s accuracy
    • Overall accuracy
    • K

Task 1 Model 3

  • Use Subtractive Clustering
  • Class Dependent for clusterInfluenceRange = 0.7
  • Change the output function to constant
  • Train the TSK model with hybrid method (Backpropagation and Least Squares Method)
  • Evaluate the model
    • Error matrix
    • Producer’s accuracy – User’s accuracy
    • Overall accuracy
    • K

Task 1 Model 4

  • Use Subtractive Clustering
  • Class Dependent for clusterInfluenceRange = 0.9
  • Change the output function to constant
  • Train the TSK model with hybrid method (Backpropagation and Least Squares Method)
  • Evaluate the model
    • Error matrix
    • Producer’s accuracy – User’s accuracy
    • Overall accuracy
    • K

Task 2

  • Use Subtractive Clustering
  • Create a grid search for the best number of features and values of radii
  • Use relieff for feature selection
  • Compare metrics between models to find the best parameters

About

TSK models using the hybrid method for classification problems

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages