Logo.gif (10726 bytes)

MasterMiner™ Product Demo

|Home|Products|Application Notes|Contact|

Demo Download & User's Instructions

Download Instructions:
1) click here and double click on file name masterminer.zip (4 MB)
2) use Winzip6.3 or higher (or evaluation version) to extract all the files of masterminer.zip
3) double click on file setup.exe and follow through. By default, the program will be installed in
C:/program Files/Mdi/.  The program is "mdi.exe" or "demo2.exe," depending on the release version
4) After starting the program, select Demo option of the FILE menu to run the 3 sets of sample data separately:
     ZZ72 - predict optimal leaching rate in aluminum production (explained below)
     YGF - predict a new phosphor from
     VPTC (vaporized positive temperature coefficient) - ceramic semiconductor design.


User's Instructions
(Steps in Testing ZZ72.dat)

Note: in the following, "X:Y:Z"means select option Z under Y sub-menu of the X menu), and CWAV stands for "Close the window after viewing it)."

  1. Start the "mdi.exe" "or demo.exe" program. Select File:Demo to open the demo data file "ZZ72.dat." A window pops up with a data table where: No. is the sample number, "kind" the class (1 or 2) of the sample, "t" the target (output) to be optimized, and "a1" through "a4" the features (inputs, factors). The objective of this project to build a model that can be used to optimize the "t."

  2. Select Statistics:Data Type. a window pops up with with 2 PRESS (prediction residual error squared sum) values and "inclusion type." Close this window after viewing (CWAV).

  3. Select Statistics:Data PLS. a window pops up prediction result by PLS (partial LS). CWAV.

  4. (Factor analysis, correlation) Select Picture:Bi-Variants:Target-Feature. Click OK on the pop-up window. A "Target-Feature" Diagram appears showing the t-a1 relationship graph. Hit the space key to view the graph of t-a2, t-a3, and t-a4 separately. Close all windows except for the ZZ72.dat window.

  5. (Factor analysis, association) Select Select Picture:Bi-Variants: Feature-Feature. Click OK on the pop-up window. A "Feature -Feature" Diagram appears, showing the a1-a2 relationship. Hit the space key to continue viewing the diagram for a1-a3, a1-a4, a2-a3, a2-a4, and a3-a4, separately. Close all windows except for the ZZ72.dat window.

  6. Select Picture:PCA (or Fisher), click OK to view the data separation result by the PCA (principal component analysis) method. A few windows pop up. No data separation is achieved by PCA, since the red and blue points are mixed in the red box shown in the last window. Close all windows except for that of ZZ72.dat.

  7. Select Picture:MREC (or :LMAP for linear mapping). Click OK to view the data separation by the MREC (map recognition) method.  Select Model:By PCA Method:Auto Square to view a better  separation by a second projection using PCA.  Complete separation is achieved now.

    1. Click P(1) or P(2) on the PCA Diagram window shown, an equation for the model appears.

    2. Select Draw:Auto Square:Equation to view all the equations and inequalities for the complete model describing all the red data points in the red box.

  8. Close all windows except for that of ZZ72.dat.

  9. Select Model:ANN:Train to train an artificial neural network using the data. Click OK to start training using the default settings, After about 3 min when the error curve stops moving, the training is completed.  Close the top window with "Error ... Steps". Now the window "ANN Results" displays a table where D(*) is the estimation error, and t(exp) and t(cal) are the expected and calculated (estimated) target value.

  10. Select Model:ANN:Parameter to view the various parameters of the trained ANN. CWAV.

  11. Select Model:ANN:Prediction to predict the target (t) value given a1,..., a4 values. Enter a value for a1, ..., a4, click OK to see the predicted t value. Click "Next" to perform the next prediction. CWAV.

  12. Select Model:ANN:Sensitivity to see the change of t value as a function of a feature (factor). Click OK to perform the sensitivity analysis using the default settings. Hit the space key on the key board to change to the next feature. Close Sensitivity Diagram and Sensitivity Selection windows.

  13. . Select Model:ANN:Two Color Diagrams to obtain a Fisher diagram for a2 and a4. CWAV.

  14.   Close all windows except for the z72.dat window.

  15. . Select Picture:MREC, try Draw:Line, Draw:Regions to draw a line of region in the 2-d space to achieve better data separation when combined with MREC method.

  16.   Steps for testing the other two sets of sample data (YGF.dat, VPTC.dat)are similar, but request special skills and knowledge due to the high non-linearity and noise level in data (for advanced user only).
       

Back to Home

Email questions to support@zaptron.com.
© 1998-2000 ZAPTRON Systems, Inc.