I am relatively new to neural networks and I have tried to use the Neural Network pattern recognition app in Matlab. I am not sure if I was successful in my approach since I I always get 100% accuracy in my confusion matrix when I use NN pattern recognition app.
I am also curious to know how to get the number of hidden layers to have better accuracy, use different learning algorithm etc.
This is the problem statement
I have a pressure sensor and depending on the pressure applied resistance is measured between 2 points of the 3 electrodes connected to it.
That is either between electrodes
1 and 2 with 2 being ground or
2 and 3 with 3 being ground or
3 and 1 with 1 being ground The pressure is applied at 2 positions on the sensor (position A and B)
[the points of pressure are chosen at random within A or B during a measurement cycle. which means it remains same during a cycle but changes in the next cycle]
This are the set of measurements I had done
1. Measured R12, R23, R13
(12(ground) No Pressure)
(23(ground) No Pressure)
(31(ground) No Pressure)
2. While applying pressure on random point on A, measured R12, R23, R13
(12(ground) Pressure on A )
(23(ground) Pressure on A )
(31(ground) Pressure on A )
3. While applying pressure on random point on B, measured R12, R23, R13
(12(ground) Pressure B )
(23(ground) Pressure on B )
(31(ground) Pressure on B )
This setup was run for 52 cycles
I want to make an NN pattern recognition program in which depending on the resistance inputs given the program should tell me if the input was from position A or B or when no pressure was applied to the sensor.
This is how i have arranged the data in nmatlab
3 rows are electrode 12,23 and 31
first 1560 columns (52 cycles *30 measurements) = no pressure data
second 1560 columns = pressure on A data
Third 1560 columns = pressure on B data
first 1560 columns = no pressure data 
second 1560 columns = pressure on A data 
Third 1560 columns = pressure on B data 
I get 99.2% accuracy when i give H value as 4 (epoch=81;performance 0.0037) and on giving h=5 (epoch=223;performance 1.8673e-07) i get 100%.
The more I increase the h value after 5, I get the same 100%. I am not sure where I must stop.
I am not sure if the issue is with the logic follow or if the data is very good.
I am attaching the program with the excel sheet where the key to the arrangement of the data is shown.
Any form of help is appreciated.
7 pekerja bebas membida secara purata kr307 untuk pekerjaan ini
------------------- Neural Network/Image processing Professional ----------------------- Relevant Skills and Experience Machine Learning, Matlab and Mathematica, Neural Networks Proposed Milestones kr212 SEK - 1