Table $3$ Original experimental dataThe calibration data of the $S{O}_2$ sensor is shown in Table $1.($Original experimental data is shown

in Table $3)$Use the least squares method to calculate nonlinearity, and calculate hysteresis error

and repeatability error.

Table $1$ Sensor response to $S{O}_2$

Then, try to optimize data by using a neural network model

$(1)$ Open the "Application" in Matlab and find the Neural Network Toolbox; You can also enter

"nnstart" directly on the command line, which will automatically jump out of the toolbox of the

neural network for you to choose.

$(2)$ Design a BP neural network program.

$(3)$ Set the network parameters. It is recommended to use the function "newff" to create a BP

neural network, the implicit transfer function is recommended to use the Sigmoid function, and

the output layer is recommended to use pure linear function.

$(4)$ Substitute training samples. Train the created BP neural network with the function train.

After $1000($recommended$)$ iterations.

$(5)$ Fill in the BP neural network model structure parameters.

Table $2$ BP neural network structure parameters of $S{O}_2$ sensor

$(6)$ The hysteresis error, repeatability error, and nonlinearity are recalculated using the trained

BP neural network and compared to the metrics previously calculated by the least squares method.