Dear Khaled,

Based on my experience it is not sufficient for developing a model based on ANFIS. It is too little.

Dear bahrouz,

I have 140 data points, which means 20 values for each variable (20*7=140)

Thank you very much Mr Bahrouz Ahmadi, I will try with PCA.

Peace be upon you

Even with a restrained dataset you can use the expert insights to help you to calibrate the ANFIS

Dear Khaled, separate your samples into training (106 Sample) and Test (34 Sample). Then you need to choose an ANFIS structure with less number of parameters (premise+consequent) than your training sample. For example, choose Gaussian-shaped with three MFs for one input, two MFs for two inputs and one MF for four input. By this structure you have 106 unknown parameters and 106 training data.

Dear Mohhammad

Thank you very much Mr Mohammadjavad Zeinali