Indoor localization and monitoring are applied using RFID technique, for which the RFID reader and tags are distributed as shown in Figure $1.$ The figure illustrates how a person can be monitored using $8$ vertical RFID sensors mounted on the wall of a room. Remember there are no tags mounted on the person. In this example, four types of person orientations are identified and these are:

$1.$ The person lying on the floor.

$2.$ The person lying on the bed

$3.$ The person sitting on a chair.

$4.$ Standing person.

Figure $1.$ Wall$-$mounted RFID tags for a localized and monitoring person in an indoor environment.

Data are collected from $8$ tag elements including the target data that identified the type of the person orientation. These are two matrices $$t$$ and $$x$$ represent the target and the output data respectively.

Three data files will be supplied to study the performance of the system in terms of neural network training confusion. These data files are named as follows:

$$ eight$\_$sensors$\_1.$mat

$$ eight$\_$sensors$\_2.$mat

$$ eight$\_$sensors$\_3.$mat

For each data set, you have to comment on the results and illustrate the following:

$1.$ Write your MATLAB programme and

$2.$ Plot the data for each case. Try to plot all output data from all $8$ sensors$.$

$3.$ Comment on the confusion matrix and errors for each data file using various sizes of the hidden layer such as $5,10$ and $20.$

$4.$ For each data file, state the variations on the performance and the ROC.

$5.$ What are the benefits of such a technique compared to wearable sensors applications?

$6.$ Create the classification tree for the three data set, remember to transform vectors $($actual data $$t$$ vector$)$ to indices Classes using the matlab function vec$2$ind$($t$)$ so that can be applied to the fitctree function and then comment on the results achieved. You have to add your pseudo$-$code on the main matlab window into your answer.

$7.$ Train four different classifiers $($namely: $\{\text{'}$Naive Bayes', 'Discriminant Analysis', 'Classification Tree', 'Nearest Neighbor'$\})$ for two data sensors $1$ and $5$ only of one data set above. Create a grid of points spanning the entire space within some bounds of the actual data values, and then predict the patient orientation using all classifiers. Comment on your observation.

$8.$ Samples of output data from the $8$ sensors will be provided to you in a file named $$collected$\_8\_$sensors$.$mat$.$ Can you provide the estimated target data for these samples? Remember to try this on the trained data of the three files given above to you.