Virtual Humans Forum
Virtual Humans Forum
Home | Profile | Register | Active Topics | Members | Search | FAQ
Username:
Password:
Save Password
Forgot your Password?

Virtual Humans
Partner website:
Chatbots.org
 All Forums
 Virtual Humans
 Artificial Intelligence
 How to predict the weather on following day
 New Topic  Topic Locked
 Printer Friendly
Author Previous Topic Topic Next Topic  

myra
Curious Member



1 Posts

Posted - May 29 2013 :  16:27:28  Show Profile
Hi. Im AI student and still new. I would like to request assistance from anybody to teach me how to predict the weather on following day such problems in the link. Really need help :( Im not sure how to get the answer. Btw, this modul don't have tutor & our lecturer quite busy with other things.

http://imageshack.us/photo/my-images/827/photozml.jpg/

mikmoth
Moderator



USA
2077 Posts

Posted - May 29 2013 :  19:44:08  Show Profile  Visit mikmoth's Homepage
You could use a neural net to predict what the temperature will be the next day. But you would need a lot of past data to make it even a little accurate... and as we know the weather is so unpredictable it's like predicting the lottery.

I was using an open-source library for my neural nets called FANN. It's pretty good and written in C++.

http://leenissen.dk/fann/wp/

If you don't know how to program then you might want to look up something simpler like a VB.Net project of a neural net.

Really... you at least have to know a little programming to make this work. Even for us who know how to program... this is not an easy project to create.

Good luck!

 http://lhandslide.com
Go to Top of Page

hologenicman
Moderator



USA
3322 Posts

Posted - Jun 04 2013 :  19:58:04  Show Profile  Visit hologenicman's Homepage

I don't have a link for this, but there was a past experiment with predicting the weather that I remember.

They took a whole year's worth of weather maps for a specific airport and fed them into a neural net. They allowed the neural net to try predicting the next day's weather and then either punished or rewarded the neural net's connections according to it's wrong or right prediction. Since they had access to the actual weather for the year, the programers/teachers had omnipotent knowledge of the weather for that whole year.

It turned out that they considered the experiment a failure since the neural net only managed to be able to predict the weather with 80% accuracy.

The reason that I remember this experiment is because the hillarious part is that the real life human weather forcaster was only able to predict the weather with a 75% accuracy at the same airport...

This experiment was done over 25 years ago, and now my local news weather forcast includes a "future Cast" which graphically shows the predicted course of weather over an extended period of time. Kool.

For your project, it would seem that your neural net only has 4 inputs, and I agree with Mikmoth that a neural net would be a good solution, however, you only have a limited number of iterations given with witch to "train" your net.

John L>

HologenicMan
John A. Latimer
http://www.UniversalHologenics.com

"If the Human brain were so simple that we could understand it,
we would be so simple that we couldn't..."
-Emerson M Pugh-

Current project:http://www.vrconsulting.it/vhf/topic.asp?TOPIC_ID=816&whichpage=1

DISCOVERY: The more I learn, the more I learn how little I know.
GOAL: There's strength in simplicity.
NOTE: Goal not always achieved.
Go to Top of Page

hologenicman
Moderator



USA
3322 Posts

Posted - Jun 04 2013 :  20:22:27  Show Profile  Visit hologenicman's Homepage
Sorry,

to make your neural network more efficient you should build a network with a temporal cascade that allows for the progression of weather over time. It's not as complicated as it sounds:

14 days worth of data with 4 inputs per day. Lets shorthand reference to the data points to (DAY,Condition)and use numbers for the days label and letters for the condition. Day 5 temperature would be (5,B).

Let's say that you cascade 4 days(the more days the better...):

First iteration
Input Array:
(1,A)(1,B)(1,C)(1,D)
(2,A)(2,B)(2,C)(2,D)
(3,A)(3,B)(3,C)(3,D)
(4,A)(4,B)(4,C)(4,D)
Output Array:
(5,A)(5,B)(5,C)(5,D)

Take this output array and back-propagate the network according to it's right and wrong answers of the actual day 5 weather conditions then run through the second iteration.

Second iteration
Input Array:
(2,A)(2,B)(2,C)(2,D)
(3,A)(3,B)(3,C)(3,D)
(4,A)(4,B)(4,C)(4,D)
(5,A)(5,B)(5,C)(5,D)
Output Array:
(6,A)(6,B)(6,C)(6,D)

Take this output array and back-propagate the network according to it's right and wrong answers of the actual day 6 weather conditions then run through the third iteration.

third iteration
Input Array:

(3,A)(3,B)(3,C)(3,D)
(4,A)(4,B)(4,C)(4,D)
(5,A)(5,B)(5,C)(5,D)
(6,A)(6,B)(6,C)(6,D)
Output Array:
(7,A)(7,B)(7,C)(7,D)

Take this output array and back-propagate the network according to it's right and wrong answers of the actual day 7 weather conditions then run through the fourth iteration.

You see that the data is temporally(time) cascaded(progressed) as the iterations progress. By the time you get to the iteration that has day 15 as the output, you should have the network fairly trained to give a relatively accurate weather prediction for day 15.

True artificial intelligence must allow for the passage of time since all consciousness exists in the passage in time.

Have fun with your project,

John L>

HologenicMan
John A. Latimer
http://www.UniversalHologenics.com

"If the Human brain were so simple that we could understand it,
we would be so simple that we couldn't..."
-Emerson M Pugh-

Current project:http://www.vrconsulting.it/vhf/topic.asp?TOPIC_ID=816&whichpage=1

DISCOVERY: The more I learn, the more I learn how little I know.
GOAL: There's strength in simplicity.
NOTE: Goal not always achieved.

Edited by - hologenicman on Jun 04 2013 20:26:13
Go to Top of Page
  Previous Topic Topic Next Topic  
 New Topic  Topic Locked
 Printer Friendly
Jump To:
Virtual Humans Forum © V.R.Consulting Go To Top Of Page
This page was generated in 0.14 seconds. Snitz Forums 2000