i am using one year waveheight data of 2003 from jan to dec with 30 minute interval and trying to predict april month using jan,feb,march datasets using feed forward nueral networks but i am getting coefficient of regression value as 0.44

[xn,xmean,xstd,yn,ymean,ystd]=prestd(x,y);

net=newff(minmax(xn),[15 1],{'tansig','purelin'},'trainlm');

[net,tr]=train(net,xn,yn);

an=sim(net,xn);

a=poststd(an,ymean,ystd);

p=new input;

q=new output;

pn=trastd(p,xmean,xstd);

on=sim(net,pn);

o=poststd(on,ymean,ystd);

[m,n,r]=postreg(o,q);

this is my code .i am getting coefficient of corelation as 0.423

[John D'Errico]

So? Do ALL models need to fit any data at all well? If all that mattered was that you want it to fit better, that is the mathematical equivalent to merely clicking a pair of ruby red shoes together to solve all problems.

[patchava mahendra]

then how to solve this wate data problem using nueral networks?is there any alternative

[patchava mahendra]

do you means we can't solve this data set using nueral networks?