Suppose a neural network with its inputs. But some of this inputs have correlation. And has a cause effect relation. Is there any way to find these inputs?
Hi,
1- you can check the inputs correlation with calculating correlation coefficients matrix (corrcoef)
2- Employ PCA to choose the inputs with least possible correlations (order reduction)
3- Employ PCA to reduce the correlation of inputs.
You may whitening method to withe the inputs before using them. PCA is just one solution of whitening .
Principal component analysis (PCA) transforms the input data so that the elements of the input vector set will be uncorrelated. In the situations that the dimension of the input vector is large, but the components of the vectors are highly correlated (redundant). It is useful in this situation to reduce the dimension of the input vectors.
An effective procedure for performing this operation is principal component analysis. This technique has three effects:
processpca command in matlab process columns of matrix with principal component analysis
Dear Pooria,
I suggest you to use V-WSP algorithm in any purpose of finding the correlated variables. Its results are awesome as variable reduction tool. More details about how to perform it and further explanations can be found in the following article:
http://www.sciencedirect.com/science/article/pii/S0169743914001154
Regards,
Be aware that PCA will project your original set on space which combine linearly your original features. If you which to still keep you original features, you need to apply features selection method and not dimensionality reduction. For that you can check the statistical test (t-test, etc.) or something like minimum redundancy maximum relevance (mRMR). Thus, you will select the features the most discriminative before to feed the ANN.
you can use the PERSON correlation function...
you can have formulea
https://www.researchgate.net/publication/269106969_A_Novel_Recurrent_Neural_Network_Model_A_Case_Study_in_Energy_Load_Forecasting
Thesis A Novel Recurrent Neural Network Model: A Case Study in Ener...
I agree with Guillaume: you should choose input variables according to its influence for modelling your objective task using the trained neural network models.
TO FIND THE IMPORTANCE OF THE INPUTS ON THE OUTPUT THE MOST EASY TO USE METHOD IS THE COSINE AMPLITUDE METHOD (CAM)
- Badges
- Science topic