Share
Google Translate
Recent blog posts
- Indonesian Government will conduct 2010 Population Census
- PDF and Online Manual Documentation of Zaitun Time Series version 0.2.1 are Now Available
- Zaitun Time Series Participated in APICTA 2009
- Zaitun Time Series version 0.2.1 is released
- Help Zaitun Time Series to participate The Asia Pacific ICT Awards (APICTA) 2009 in Melbourne
User login
Neural Network Analysis with Zaitun Time Series
Thu, 11/20/2008 - 21:24 — admin
Zaitun Time provides neural network modeling of time series data. To perform neural network modeling on a time series variable:
- Click Analysis -> Neural Network
- Select Variable Dialog appears. Choose a variable you want to build its neural network model, and then click OK.
- The Neural network analysis form will appear. Determine the parameters of the neural network model you want to build. You can determine the parameters of the neural network architecture, activation function, and learning algorithm. You can also set up the stopping condition or use early stopping cross-validation method.
- Click Start. The learning process will start and run until the stopping condition is fulfilled or the operation has reached the maximum number of iterations.
- You can stop the learning process any time by clicking Stop while the learning process is running.
- After the learning process is finished, click View Result button to display the model result.
- The Select Result View dialog will appear. Select the result views you want to display, and then click OK. You can forecast the data by clicking the Forecasted item and determine the number of data you want to forecast.
- The selected model result will be displayed on Result View panel in several tabs.
Comments
Salam,
I'd like to ask a question on what I'm doing with neural network model estimation.
I've some theoretical background on the structure of neural models, and planing to use it for forecasting inflation. I'll use a simple autoregressive model of inflation, in which current inflation is used as output and its lagged values as inputs. and then compare the forecasting accurracy with the linear models.
can you explain functional form of the model estimated with zaitun and also give some information what are the input layers? are they lag or lead values of output layer?
Thanks
Salam
Thank you for your question.
In Zaitun Time Series we use Fully Connected Feed Forward Neural Network topology with Backpropagation learning algorithm.
The model used in Zaitun Time Series is simple Autoregressive (AR) which is:
Yt = f(Yt-1,Yt-2,Yt-3,...Yt-n)
which n means the number of input layer neuron.
So, for example, if the number of input neuron is set to 6 the model is:
Yt = f(Yt-1,Yt-2, Yt-3,Yt-4,Yt-5, Yt-6)
The values of Yt-1,Yt-1,Yt-2, Yt-3,Yt-4,Yt-5, Yt-6 will be the value of each 6 neurons in input layer, and the value of Yt will be the value of output layer neuron.
Hope it will answer your question.
If you need more explanation or have another question just give a reply comment on this comment.
Salam.
Thank you
Rizal Zaini Ahmad Fathony
Zaitun Time Series Developer Team
Post new comment