ocsvmnode properties

One-Class SVM node iconThe One-Class SVM node uses an unsupervised learning algorithm. The node can be used for novelty detection. It will detect the soft boundary of a given set of samples, to then classify new points as belonging to that set or not. This One-Class SVM modeling node in SPSS Modeler is implemented in Python and requires the scikit-learn© Python library.

Table 1. ocsvmnode properties
ocsvmnode properties Data type Property description
custom_fields boolean This option tells the node to use field information specified here instead of that given in any upstream Type node(s). After selecting this option, specify the fields below as required.
inputs field List of the field names for input.
role_use string Specify predefined to use predefined roles or custom to use custom field assignments. Default is predefined.
splits field List of the field names for split.
use_partition Boolean Specify true or false. Default is true. If set to true, only training data will be used when building the model.
mode_type string The mode. Possible values are simple or expert. All parameters on the Expert tab will be disabled if simple is specified.
stopping_criteria string A string of scientific notation. Possible values are 1.0E-1, 1.0E-2, 1.0E-3, 1.0E-4, 1.0E-5, or 1.0E-6. Default is 1.0E-3.
precision float The regression precision (nu). Bound on the fraction of training errors and support vectors. Specify a number greater than 0 and less than or equal to 1.0. Default is 0.1.
kernel string The kernel type to use in the algorithm. Possible values are linear, poly, rbf, sigmoid, or precomputed. Default is rbf.
enable_gamma Boolean Enables the gamma parameter. Specify true or false. Default is true.
gamma float This parameter is only enabled for the kernels rbf, poly, and sigmoid. If the enable_gamma parameter is set to false, this parameter will be set to auto. If set to true, the default is 0.1.
coef0 float Independent term in the kernel function. This parameter is only enabled for the poly kernel and the sigmoid kernel. Default value is 0.0.
degree integer Degree of the polynomial kernel function. This parameter is only enabled for the poly kernel. Specify any integer. Default is 3.
shrinking Boolean Specifies whether to use the shrinking heuristic option. Specify true or false. Default is false.
enable_cache_size Boolean Enables the cache_size parameter. Specify true or false. Default is false.
cache_size float The size of the kernel cache in MB. Default is 200.
enable_random_seed Boolean Enables the random_seed parameter. Specify true or false. Default is false.
random_seed integer The random number seed to use when shuffling data for probability estimation. Specify any integer.
pc_type string The type of the parallel coordinates graphic. Possible options are independent or general.
lines_amount integer Maximum number of lines to include on the graphic. Specify an integer between 1 and 1000.
lines_fields_custom Boolean Enables the lines_fields parameter, which allows you to specify custom fields to show in the graph output. If set to false, all fields will be shown. If set to true, only the fields specified with the lines_fields parameter will be shown. For performance reasons, a maximum of 20 fields will be displayed.
lines_fields field List of the field names to include on the graphic as vertical axes.
enable_graphic Boolean Specify true or false. Enables graphic output (disable this option if you want to save time and reduce stream file size).
enable_hpo Boolean Specify true or false to enable or disable the HPO options. If set to true, Rbfopt will be applied to find out the "best" One-Class SVM model automatically, which reaches the target objective value defined by the user with the following target_objval parameter.
target_objval float The objective function value (error rate of the model on the samples) we want to reach (for example, the value of the unknown optimum). Set this parameter to the appropriate value if the optimum is unknown (for example, 0.01).
max_iterations integer Maximum number of iterations for trying the model. Default is 1000.
max_evaluations integer Maximum number of function evaluations for trying the model, where the focus is accuracy over speed. Default is 300.