Critical Condition Detection Using Lion Hunting Optimizer and SVM Classifier in a Healthcare WBAN

A﻿timely﻿critical﻿condition﻿detection﻿and﻿early﻿notification﻿are﻿two﻿essential﻿requirements﻿in﻿a﻿healthcare﻿wireless﻿body﻿area﻿network﻿for﻿the﻿correct﻿treatment﻿of﻿patients.﻿However,﻿most﻿of﻿the﻿systems﻿have﻿ limited﻿capabilities﻿and﻿so﻿could﻿not﻿detect﻿the﻿exact﻿condition﻿in﻿a﻿precise﻿time﻿interval.﻿In﻿addition﻿ to﻿these﻿it﻿needs﻿a﻿reduction﻿in﻿the﻿false﻿alert﻿rate,﻿as﻿issuing﻿alerts﻿for﻿the﻿deviation﻿in﻿each﻿incoming﻿ packet﻿increases﻿the﻿false﻿alert﻿rate﻿and﻿these﻿false﻿alerts﻿consume﻿more﻿network﻿resources.﻿In﻿order﻿ to﻿fulfill﻿the﻿above-mentioned﻿requirements,﻿a﻿dynamic﻿alert﻿system﻿has﻿been﻿designed﻿in﻿this﻿regard﻿ to﻿make﻿it﻿more﻿efficient,﻿also,﻿a﻿new﻿kind﻿of﻿hybridization﻿approach﻿is﻿being﻿introduced﻿to﻿it﻿with﻿ the﻿additive﻿support﻿of﻿a﻿nature-inspired﻿optimization﻿strategy﻿named﻿Lion﻿Hunting﻿and﻿a﻿machine-learning﻿technique﻿called﻿support﻿vector﻿machine.﻿The﻿simulation﻿is﻿done﻿using﻿a﻿network﻿simulator﻿ NS-2.35,﻿and﻿the﻿proposed﻿alerting﻿system﻿outperforms﻿others.


RELATED WORK
In this section, a brief overview on the work that has been done using LH and SVM described separately.Afterthat,adetailedstudyoftheseworkswhichhadbeentakenplacetilldateinregards toanalertingsystemusingahybridizationconceptofbothnature-inspiredandmachinelearning techniquesareexplained.
The workflow diagram for whole LH-SVM based alerting system is explained in Figure 1.Algorithm1presentsthepseudo-codefortheproposedalertingsystem.Received a packet 4.
Store value of AD in CN's database 8.
Do not activate the critical_field this packet of t i.e. critical_field=1 12. Until timer expired 13.Read all AD values from the database 14.Sort them into ascending order 15.Estimate the weighted standard deviation (WSD) using Equation ( 13 Estimate the position of the hunters according to the position of victim using Equation ( 9)-(10) 7.
If current position of hunters are better than their previous position Then 8.
They may capture the victim 9.
The victim may escaped and obtained a new potion 11.
Estimate the new position of the victim using Equation (8) 12.

PERFORMANCE EVALUATION
The proposed model is evaluated by means of above study to differentiate the performance gap betweenSAD,DASanditself.TheeffectivenessoftheproposedLHSVMASisactuallyimproved becauseofcombinedapplicationofbothLHoptimizationandSVMclassification.Hereinorderto judgetheperformancecapabilityofeachonevariousQoSmetricsi.e.loss,delay,andthroughput arecomparedandevaluated.ThegeneratedresultsshowtheefficiencyofbothLHandSVMinthis regards.Theimplementationofallthesethreealertingsystemsareconductedinthens-2.35networksimulatorwhichevaluatestheirpotentialwithrespecttovariousQoSmetricsandaMATLABtool is used to generate the performance evaluations graphs for each QoS metrics.The simulation is carriedoutforanincreaseinthenumberofsensornodesothattheefficiencyofeachsystemcanbe measuredinhighaswellaslowload.

Figure
Figure 2. Performance evaluation graph for packet loss ratio

Figure
Figure 3. Performance evaluation graph for transmission delay whichdecreasesthelocalsolutionandhelpsto findtheglobalsolutionwithinlesstime.
Randomly generate a population of hunter lioness 3.According to their hunting ability segregate these lions into three sub-groups i.e. middle, left and right 4. Estimate the initial position of the victim or the prey using Equation (7) k 2.