Consider the following data set where there are five input attributes and one output

$($i$.$e$.,$ Diagnosis$).(20$ points$)$

Gender Height Weight Age Intensity of

Symptoms

DIAGNOSIS

Female Tall High Old High Sick

Female Short High Young Low Healthy

Male Short High Young Low Healthy

Male Short Low Old High Sick

Male Short Low Middle Low Sick

Female Short Low Young High Healthy

Female Tall High Middle Low Healthy

Male Tall Low Young Low Sick

Female Tall High Middle High Healthy

Female Tall Low Young High Healthy

Suppose that you want to choose the best subset of features, which would yield the highest accuracy

when Na$$ve Bayes classifier is used.

After some thought you realized that even though due to the small size of the problem brute force

approach $($i$.$e$.,$ total enumeration$)$ would easily provide you the result, this would be a good

opportunity to improve your knowledge in metaheuristics and decided to implement simulated

annealing algorithm as the search algorithm that would lead you to a good solution.

Assume $[10010]$ as the starting solution and top priority upfront as the neighborhood generation

mechanism. What would be the next solution if training error was used as the cost function and first

improve strategy was adopted. Assume temperature $=100.$

Assume the following random numbers $($in order$)$ in case you need to use them in your answer:

$0,236,0\mathrm{.}502,0\mathrm{.}752,0\mathrm{.}101,0\mathrm{.}348,$ and $0\mathrm{.}756.$