Data overfitting is a common problem in FIS parameter optimization. When overfitting occurs, the tuned FIS produces optimized results for the training data set but performs poorly for a test data set. Due to overturning, the optimized FIS parameter values pick up noise from the training data set and lose the ability to generalize to new data sets. The difference between the training and test performance increases with the increased bias of the training data set. To overcome the data overfitting problem, a tuning process can stop early based on an unbiased evaluation of the model using a separate validation dataset.
How to do such Parameter optimization of a fuzzy inference system (FIS) ?
I'm not familiar with FIS systems. But speaking of optimizing parameters, perhaps you can use some fine-tuning strategies. Basically,
- if your model has hyper-parameters that are not directly learned within estimators, you can consider using Grid search, have a look at https://scikit-learn.org/stable/modules/grid_search.html.
- To optimize the normal parameters of models, just do cross-validation based on your measure metrics to select the best parameter setting.
- Also, you can use feature select approaches to pre-process your data before training your model, which could give better results.
Hopefully, it helps you somehow.
Have a look at Article An Overview of Evolutionary Algorithms for Parameter Optimization
https://hal.archives-ouvertes.fr/hal-01320328/document
Conference Paper Parameter optimization of a Fuzzy Inference System using the...
k-fold cross validation is a re-sampling procedure to assess the performance of classification models used.
However, to for parameter optimization various optimization algorithm can be used such as GridSearch, RandomSearch, and Swarm and Evolutionary Algorithms.
The following link may help:
https://www.mathworks.com/help/fuzzy/tune-fuzzy-inference-systems.html
- Badges
- Science topic
- Similar topics
- Organizational Psychology
- Training