Support Vector Machine with Deployment (Regression)

Full-featured implementation of Support Vector Machines (SVM) for regression problems. The final solution is automatically stored for deployment.

General

Element Name Description
Detail of computed results reported Detail of computed results; if Minimal detail is requested, spreadsheets of analysis summary, model specifications as well descriptive statistics (regression statistics) will be displayed; at the Comprehensive level of detail, a spreadsheet of predictions and residuals as well as their histogram plots will be displayed; in addition to the above, the All results level will display a spreadsheet (if the 'Creates residual statistics' option is selected) containing all data set variables and their statistics including predictions and residuals/accuracy (whichever applicable).
Missing data deletion Specifies the substitution method for missing data. Casewise excludes cases that contain any missing data for any of the selected variables in the analysis. Mean substitution replaces missing data by the means for the respective variables (Note: This option is not applicable for categorical dependent and predictor variables).
Generate datasource, if N for input less than Generate a data source for further analyses with other Data Miner nodes if the input data source has fewer than k observations, as specified in this edit field; note that parameter k (number of observations) will be evaluated against the number of observations in the input data source, not the number of valid or selected observations.

Sampling

Element Name Description
Divide data into train and test samples Divides the data set into training and test sample. The training subset is used to fit the model while the test subset serves as an independent check of its performance.
Sampling method Sampling method to be used for dividing the data set into train and test subsets. Random sampling will divide the data set into train and testing samples in a random fashion. This is in contrast to the First N method which selects the first N cases as the training set and the rest as the testing sample. NOTE: you may also use a learning/testing indicator variable method for sampling from the data. You can access this functionality via the Advanced tab of the data spreadsheet in the Data Acquisition of Statistica Data Miner environment. Selecting this method (i.e. learning/testing indicator) will override any choice of sampling you make on this tab.
Size of training sample Specifies the percentage of data cases that will be used to form the training sample. The remaining valid cases in the data set will be used as the test sample.
Seed Specifies the random generator seed for random sampling of data into train and test subsets.
Use first N cases Selects the first N valid cases in the data set as training subset. The rest are used for testing.

SVM

Element Name Description
SVM type Specifies the type of the SVM model.
Capacity Capacity parameter.
Epsilon Epsilon parameter.
Nu Nu parameter.

Kernel

Element Name Description
Kernel type Specifies the type of the Kernel used by the SVM model.
Degree Specifies the degree of the polynomial kernel.
Gamma Specifies the gamma parameter for polynominal, RBF and sigmoid kernels.
Coefficient Specifies the coefficient for polynominal and sigmoid kernels.

Cross-validation 1

Element Name Description
Apply v-fold cross-validation Applies v-fold cross-validation to obtain estimates of the capacity, epsilon and nu parameters.
V value Number of cross-validation folds.
Seed Seed value for random data shuffling for cross-validation.
Minimum C Start value for the capacity parameter (used by cross-validation grid search).
Maximum C End value for the capacity parameter (used by cross-validation grid search).
Increment in C
Minimum Epsilon Start value for the epsilon parameter (used by cross-validation grid search).
Maximum Epsilon End value for the epsilon parameter (used by cross-validation grid search).
Epsilon increment Increment in epsilon.

Cross-validation 2

Element Name Description
Minimum Nu Start value for the nu parameter (used by cross-validation grid search).
Maximum Nu End value for the epsilon parameter (used by cross-validation grid search).
Nu increment Increment epsilon (used by cross-validation grid search).

Training

Element Name Description
Max number of iterations The maximum number of iterations that can be applied in training the SVM model.
Stop at accuracy Training stops when the given level of accuracy is reached.
Cache size, in MB Cache size in MB
Shrink data Shrink data
Scale inputs Check this option to linearly scale the inputs within the range 0 to 1.
Scale outputs Check this option to linearly scale the outputs within the range -1 to 1.

Results

Element Name Description
Subset used to generate results Select the subset for which the results should be displayed.
Include inputs Includes the independent variables in spreadsheets and histograms.
Include outputs Includes the dependent variables in spreadsheets and histograms.
Include predictions Includes predictions in spreadsheets and histograms.
Include residuals Includes residuals in spreadsheets and histograms.
Creates residual statistics Creates predicted and residual statistics for each case depending on the selected level of details.

Deployment

Deployment is available if the Statistica installation is licensed for this feature.

Element Name Description
Generates C/C++ code Generates C/C++ code for deployment of predictive model.
Generates SVB code Generates Statistica Visual Basic code for deployment of predictive model.
Generates PMML code Generates PMML (Predictive Models Markup Language) code for deployment of predictive model. This code can be used via the Rapid Deployment options to efficiently compute predictions for (score) large data sets.