Sign in

Not a member? | Forgot your Password?

Search compdsp

Search tips

Free PDF Downloads

A Quadrature Signals Tutorial: Complex, But Not Complicated

Understanding the 'Phasing Method' of Single Sideband Demodulation

Complex Digital Signal Processing in Telecommunications

Introduction to Sound Processing

C++ Tutorial

Introduction of C Programming for DSP Applications

Fixed-Point Arithmetic: An Introduction

Cascaded Integrator-Comb (CIC) Filter Introduction

Discussion Groups

FFT Spectral Analysis Software

Free Online Books

See Also

Embedded SystemsFPGA

Discussion Groups | Comp.DSP | FFT extrapolation on wavelet transform of weather data

There are 8 messages in this thread.

You are currently looking at messages 1 to .


Is this discussion worth a thumbs up?

0

FFT extrapolation on wavelet transform of weather data - 2006-08-19 17:59:00

Recently i have been working on a problem regarding predicting weather
conditions based on given collected data( of temp, humidity etc..) of
past year or so. I decided to use FFT based extrapolation on the
wavelet transform of the data of one weather identifier(say
temperature). My reasons for doing so are:
1) I wanted to predict the future values of temperature based on past
variations. So i used FFT extrapolation which extrapolates the data
while keeping the fourier transform unchanged after extrapolation.This
implies that data predicted follows same variation.
2)But, The problem with this was that Fourier transform does not
consider the time of change in the values i.e., at what time the
changes occur.
3) So, I took the wavelet transform of the temp , which gives a
time-frequency variation of the data and then extrapolated the
coefficients of WT using FFT extrapolation.

Results->
When used to predict the temp for  100 days using data of one year.The
results had an accuracy of 87%.Similarly for humidity it showed an
accuracy of 85%. The prediction also showed remarkably good prediction
when it predicted a sudden change from 1605 to 100Btu/sqrft in the
global soal radiations falling on the area when the fall was from 1605
to 10.


Re: FFT extrapolation on wavelet transform of weather data - jim - 2006-08-20 09:05:00

s...@gmail.com wrote:


> Results->
> When used to predict the temp for  100 days using data of one year.The
> results had an accuracy of 87%.Similarly for humidity it showed an
> accuracy of 85%. 

So if I predict the high temperature for a given day is going to be 40
degrees and it turns out to be 34 degrees is that an accuracy of 85%?
There are places in this world where the temp never deviates from the
norm by more than 15%.

-jim











The prediction also showed remarkably good prediction
> when it predicted a sudden change from 1605 to 100Btu/sqrft in the
> global soal radiations falling on the area when the fall was from 1605
> to 10.

----== Posted via Newsfeeds.Com - Unlimited-Unrestricted-Secure Usenet News==----
http://www.newsfeeds.com The #1 Newsgroup Service in the World! 120,000+ Newsgroups
----= East and West-Coast Server Farms - Total Privacy via Encryption =----


Re: FFT extrapolation on wavelet transform of weather data - Rune Allnor - 2006-08-20 10:41:00

jim wrote:
> s...@gmail.com wrote:
>
>
> > Results->
> > When used to predict the temp for  100 days using data of one year.The
> > results had an accuracy of 87%.Similarly for humidity it showed an
> > accuracy of 85%.
>
> So if I predict the high temperature for a given day is going to be 40
> degrees and it turns out to be 34 degrees is that an accuracy of 85%?
> There are places in this world where the temp never deviates from the
> norm by more than 15%.

What temperature scale do you use? Where I live, the temperature
fluctuates between -20 and +30 over a year, the norm being +8 or so.
Does that make for a 400% variance?

Rune


Re: FFT extrapolation on wavelet transform of weather data - jim - 2006-08-20 10:51:00

Rune Allnor wrote:
> 
>
> 
> What temperature scale do you use? Where I live, the temperature
> fluctuates between -20 and +30 over a year, the norm being +8 or so.
> Does that make for a 400% variance?

I was thinking of tropical islands where the temp is pretty much the
same every day of the year. My point was I have no idea whether he was
complaining or boasting when he says 85% accuracy.

-jim

----== Posted via Newsfeeds.Com - Unlimited-Unrestricted-Secure Usenet News==----
http://www.newsfeeds.com The #1 Newsgroup Service in the World! 120,000+ Newsgroups
----= East and West-Coast Server Farms - Total Privacy via Encryption =----


Re: FFT extrapolation on wavelet transform of weather data - Jerry Avins - 2006-08-20 11:55:00

jim wrote:
> 
> Rune Allnor wrote:
>>
>>
>> What temperature scale do you use? Where I live, the temperature
>> fluctuates between -20 and +30 over a year, the norm being +8 or so.
>> Does that make for a 400% variance?
> 
> I was thinking of tropical islands where the temp is pretty much the
> same every day of the year. My point was I have no idea whether he was
> complaining or boasting when he says 85% accuracy.

I think Rune's point is more trenchant: 85% of what? One  might suppose 
that 32F is 47% of 68F. What then is the ratio of 0C to 20C?

Jerry
-- 
Engineering is the art of making what you want from things you can get.
¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯¯


Re: FFT extrapolation on wavelet transform of weather data - jim - 2006-08-20 12:51:00

Jerry Avins wrote:

> 
> I think Rune's point is more trenchant:

If you want to pick that nit then No I would disagree. You wouldn't
expect anyone to to be looking at the norm for the year to predict any
particular day's weather (unless you really were deserted on a tropical
island and you had lost all track of the seasons) If you were going to
reference a norm you would use the normal temperature for that day. My
point was that depending where in the world you were predicting daily
weather based on a years worth of data may or may not be so difficult.
But it mostly hinges on what you consider to be an accurate
prediction.   

`  The OP's 85% obviously means 85% correct predictions and 15%
incorrect. But what makes a prediction correct? I have heard it said
that if the weatherman predicted that the weather tomorrow would be the
same as the weather today predictions would be more accurate than they
are now. I have also heard broadcast weather predictors claim something
like 95% accuracy, but have never understood what they meant by that.

-jim

> 85% of what? One  might suppose
> that 32F is 47% of 68F. What then is the ratio of 0C to 20C?
>

----== Posted via Newsfeeds.Com - Unlimited-Unrestricted-Secure Usenet News==----
http://www.newsfeeds.com The #1 Newsgroup Service in the World! 120,000+ Newsgroups
----= East and West-Coast Server Farms - Total Privacy via Encryption =----


Re: FFT extrapolation on wavelet transform of weather data - 2006-08-20 14:13:00

Well my Prof said the same about about 15% being too much.i am doomed
:(
jim wrote:
> Jerry Avins wrote:
>
> >
> > I think Rune's point is more trenchant:
>
> If you want to pick that nit then No I would disagree. You wouldn't
> expect anyone to to be looking at the norm for the year to predict any
> particular day's weather (unless you really were deserted on a tropical
> island and you had lost all track of the seasons) If you were going to
> reference a norm you would use the normal temperature for that day. My
> point was that depending where in the world you were predicting daily
> weather based on a years worth of data may or may not be so difficult.
> But it mostly hinges on what you consider to be an accurate
> prediction.
>
> `  The OP's 85% obviously means 85% correct predictions and 15%
> incorrect. But what makes a prediction correct? I have heard it said
> that if the weatherman predicted that the weather tomorrow would be the
> same as the weather today predictions would be more accurate than they
> are now. I have also heard broadcast weather predictors claim something
> like 95% accuracy, but have never understood what they meant by that.
>
> -jim
>
> > 85% of what? One  might suppose
> > that 32F is 47% of 68F. What then is the ratio of 0C to 20C?
> >
>
> ----== Posted via Newsfeeds.Com - Unlimited-Unrestricted-Secure Usenet News==----
> http://www.newsfeeds.com The #1 Newsgroup Service in the World! 120,000+ Newsgroups
> ----= East and West-Coast Server Farms - Total Privacy via Encryption =----


Re: FFT extrapolation on wavelet transform of weather data - jim - 2006-08-20 14:44:00

s...@gmail.com wrote:
> 
> Well my Prof said the same about about 15% being too much.i am doomed

You might consider predicting the weather 1 year in advance. That's
probably just as easy as 100 days. Or better yet 10 years in advance.
The further out the forecast the better that 85% will look :}

-jim



> :(
> jim wrote:
> > Jerry Avins wrote:
> >
> > >
> > > I think Rune's point is more trenchant:
> >
> > If you want to pick that nit then No I would disagree. You wouldn't
> > expect anyone to to be looking at the norm for the year to predict any
> > particular day's weather (unless you really were deserted on a tropical
> > island and you had lost all track of the seasons) If you were going to
> > reference a norm you would use the normal temperature for that day. My
> > point was that depending where in the world you were predicting daily
> > weather based on a years worth of data may or may not be so difficult.
> > But it mostly hinges on what you consider to be an accurate
> > prediction.
> >
> > `  The OP's 85% obviously means 85% correct predictions and 15%
> > incorrect. But what makes a prediction correct? I have heard it said
> > that if the weatherman predicted that the weather tomorrow would be the
> > same as the weather today predictions would be more accurate than they
> > are now. I have also heard broadcast weather predictors claim something
> > like 95% accuracy, but have never understood what they meant by that.
> >
> > -jim
> >
> > > 85% of what? One  might suppose
> > > that 32F is 47% of 68F. What then is the ratio of 0C to 20C?
> > >
> >
> > ----== Posted via Newsfeeds.Com - Unlimited-Unrestricted-Secure Usenet News==----
> > http://www.newsfeeds.com The #1 Newsgroup Service in the World! 120,000+ Newsgroups
> > ----= East and West-Coast Server Farms - Total Privacy via Encryption =----

----== Posted via Newsfeeds.Com - Unlimited-Unrestricted-Secure Usenet News==----
http://www.newsfeeds.com The #1 Newsgroup Service in the World! 120,000+ Newsgroups
----= East and West-Coast Server Farms - Total Privacy via Encryption =----