I have Input and output data set and going for modelling. I am getting negative values of output from ANFIS model but in training as well as testing input there is no negative value. Tell me how to solve this problem.
It is not possible to get negative value if training and test set does not have negative values. share your data and anfis settings / fis file.
ANFIS (Adaptive Network-based Fuzzy Inference-System) is a fuzzy system set
applied in the context of adaptive networks. It was proposed by Jang (Jang, 1993).
For MATLAB users, ANFIS is implanted in the toolbox 'Neurofuzzy'
with his hybrid learning algorithm based on the backpropagation
and the least squares method. This model gives very good results approximating
non-linear functions.
D'abord il faut vérifier la cohérence des données entrées sorties, esuite adapter les paramètres relatifs à l'algorithme ANFIS implanté dans la boite à outils ’Neurofuzzy’.
I apologize to speak french
First you have to check the consistency of data inputs and outputs, esuite adapt the parameters for the ANFIS algorithm implemented in the toolbox 'Neurofuzzy'.
Good luck
Ernesto Sir,
Thanks for solution.Please tell me how to normalize data set.
Thanks and Regards,
Rohit Sharma
First, you should use ANFIS editor in MATLAB. Thus, you can see the ANFIS model results for your problem. You must be normalize your data with various normalize methods.
If parameters of network are setted incorectly, results can be negative. You must be set properly parameters of ANFIS network. First, you should use ANFIS editor in MATLAB. Thus, you can see the ANFIS model results for your specific problem with 8 input parameters. ANFIS editor cannot respond properly results. It depends on the nonlinearity of the problem. Later, you should use Genfis1, Genfis2 and Genfis3 models in MATLAB. (see to help file for Genfis1, Genfis2 and Genfis3 in MATLAB). If the results are still not good, you must be normalize your data.
You can use 80% or 90% of data set for ANFIS training, remainders for testing.
Hi Rohit Sharma,
You can use the normalize data module to transform a dataset so that the columns of the data set are on a common scale.
For example, the input dataset can contain columns with very different values, which can cause problems when you combine the values as features for modeling. By transforming the values so that they retain their distribution and their general reporting, but respecting a common scale, you can usually get better results when modeling.
You can apply standardization to a single or multiple columns in the same data set.
You can also record the steps of normalization that you set in this module as a transformation that can be applied to other datasets.
you can use Microsoft Azur to normalize the data, here is the link.
https://msdn.microsoft.com/fr-fr/library/azure/dn905838.aspx
Good Luck