Temparature data(data mining)

Tommorow temparture is predicted i.e predictive analysis or predictive datamining or Astrologist.
In this Percentage of error depends on parametrs and the data,which is known as Senstivity analysis.
SENSITIVITY ANALYSIS
J(x)=1/2E(i=1tom)[z(i)-h(x(i)](square)
Here there are two methods i.e Direct method and Replace of orginal method.
Direct method is
dj/dx=[z-h(x)  dh(x)/d(x)]=0
where z-h(x) and  dh(x)/d(x) is non linear which is solved by numerical method i.e newtons method.
Replace the original problem by appropriate problem by liner which is orginal exactly.
Consider the base point of the curve as Xc
h(x)=h(Xc+(X-Xc))
h(x)=h(Xc+delta^) where delta=X-Xc
h(x)=h(Xc)+deltadh(Xc)/d(x)    By taylors method
thus J(x)=1/2(Zx-h(x))square     here X is operating point
        J(x)=1/2[Zi-h(Xc)+h!(Xc)delta]square
       J(delta)=1/2[Gi-h(Xc)delta]square this is linear equation.
      dJ/ddelta=[Gi-h!(Xc)delta]h!(Xc)=0
     Gih!(Xc)=[h!(Xc)]square delta
    delta=Gih!(Xc)/[h1!(Xc)]square
                         which is an optimal value of delta.
If delta is 10^-3 the next iteration is differ for 3 times