RFIS: Regression-based Fuzzy Inference System

RFIS is a novel simple fuzzy inference system without explicitly defined fuzzy rules based on linear and nonlinear regressions.

This project concerns training fuzzy systems on the basis of linear and nonlinear regressions. These systems use Gaussian fuzzy sets for the inputs and linearly and nonlinearly parameterized system functions to obtain the output. The linear regression is realized by the ridge regression and the nonlinear regression by Levenberg-Marquardt algorithm. The input fuzzy sets are determined by a multi-objective genetic algorithm with a feature selection method. In the case of linearly parameterized system functions, the following methods are considered: F-test and a regression tree. In the case of nonlinearly parameterized system functions, terms from the so-called term matrix are coded in an individual and they are selected by using a genetic algorithm. The multi-criteria objective functions enable the selection of models from the Pareto fronts taking into account the compromise between model accuracy and its simplification.

The package contains an example of using the RFIS method to predict a time series based on the Box-Jenkins gas furnace data set. The data can be downloaded from: https://openmv.net/info/gas-furnace

The second example of using the RFIS method concers the problem of the evaluation of an authentication procedure for banknotes. The data can be downloaded from: https://archive.ics.uci.edu/ml/datasets/banknote+authentication

Comparison of the RFIS and ANFIS models can be found at: https://www.mathworks.com/matlabcentral/fileexchange/115775-rfis-vs-anfis-comparison-of-two-fuzzy-models?s_tid=srchtitle

Using this method, please cite as:

Wiktorowicz K., 'RFIS: regression-based fuzzy inference system', Neural Computing and Applications, 2022, vol. 34, pp. 12175–12196, DOI: 10.1007/s00521-022-07105-8

Available at:

https://www.researchgate.net/publication/359312375_RFIS_regression-based_fuzzy_inference_system

https://link.springer.com/article/10.1007/s00521-022-07105-8

To see a simple example, run:

illustrative_example

To predict a time-series for Box-Jenkins data, run:

boxjen_main.m

boxjen_optim.m

or

boxjen_main_nlm.m

boxjen_optim_nlm.m

To classify banknotes, run

banknote_main.m

banknote_optim.m