Tutorial 1: AgielNN Data Preparation

AgielNN is basically used to make a prediction of unknown result based on specific input parameters. However, AgielNN requires to be trained first before it ready to conduct any of forecasting. And to make sure that the training is going well, it is also recommended to conduct checking or testing using blind test. So, in general we need 3 kinds of data which are data training, data checking, and data forecasting. 

Data Training & Checking.
Data training and checking must have the same format, which by means of having similar number of column. And those two data are ideally coming from the same data source which is actual data from the field. For example, i want to make a prediction of parameter Z which is estimated to be a function of X, Y, and i have 120 real data from experience as you can see in this SampleANN1.XLS. For simplicity, the data can be divided in to training and checking data by portion of 75:25.  Before inputting to AgielNN, those divided data then must be copied to text editor (Notepad of other similar software) and saved as 2 separated files namely Sample1_train.txt and Sample1_check.txt.


Data for Forecasting
Since we want to forecast from 2 variables input, so the data for forecasting must have 2 column only. For example, please refers to this Sample1_Forecast.txt. Must be noted that all input data for forecast must in the range of data input used in the training. For example, if A in the training have minimum value = 0 and maximum = 100, then all A data that will be used for forecasting must lying on 0 to 100. Otherwise, the forecast might goes wrong since the neural network system is only trained for the specific range of data.

AgielNN doesn’t require normalization data since it is automatically do normalization. However, data variation should have linear variation and not exponential or logarithmic or else. For example, if you have variable Z as a function of Log(A), then it is recommended to convert A to C first, where C=Log(A), and use C as input for forecasting the Z.


Inputting Data to AgielNN
This is a simple task actually, just go to “Data Source Preparation” window, and click and you will be prompted with this kind of dialog. Fill the name of the data, let say “Data1_Training” and click OK.

A new data source will be created. After that, click the data name. And you will get an empty table on the right panel. To import table from the data that we have prepared previously, simply right click inside the table, and select “Import the File”, or just click the button “Import Table from File” below the table. Browse the Sample1_Train.txt and open. Repeat the procedure for Sample1_Check.txt and Sample1_Forecast.txt named as “Data1_Check” and “Data1_Forecast”. By doing that, now you should have 3 data source, Data1_Check which has 3 columns, Data1_Train 3 columns, and Data1_Forecast 2 columns and this kind of interface.

So, data preparation is finished. This should be a very easy data preparation.. 🙂

This is the end of this tutorial. Should anybody have question, please don’t hesitate to contact my email.
Written by Sugiyanto Suwono
Bekasi, March 13, 2013

3 thoughts on “Tutorial 1: AgielNN Data Preparation

    1. Actually for data preparation, no parameters need to be customized.
      If you have data from Ms. Excel, just select all data and copy then paste into AgielNN data editor.
      The format can be Ms.Excel, or simply ASCI text file with tab delimited data.

  1. Can the forecasting be done by using more than 2 input variables? For example what if I want to use 4 input variables.

Leave a Reply

Your email address will not be published. Required fields are marked *