Can anyone help me how to solve following doubts
i. Is there any signal processing should be done after simulating waveform and before supplying input to neural network.
ii. How to arrange samples in matrix form such that it would be train and tested by neural network.
Dear Rajesh, My initial thoughts are as follows:
1. Signal processing techniques aid in analyzing the variable under investigation. Considering power transformer to be the focal machine, current and flux can be decomposed using wavelet and EMD.
2. As far as classification and regression tasks are concerned, input matrix contains relevant features for your target variable (current/flux). Several other parameters like line voltage, oil temperature as a time-series can aid your ML task. Usually 70% of data is trained ans rest is tested. Usually, individual inputs can act as column vectors.
this process before it consider as preprocessing step then take the new vector to apply the waveform to get good result
Dear L. Rajesh Reddy
In my opinion, because you are dealing with some signal sets, you should process them (pre-processing), extract useful features (feature extraction- depending on your application), consider some labels (creating data), and then construct dataset. After that, you can apply a feature set to the neural network predictor or classifier.
- Badges
- Science topic