6.2 About Model Settings
You can specify settings that affect the characteristics of a model.
Some settings are general, some are specific to an Oracle Machine Learning function (also referred to as a technique), and some are specific to an algorithm.
All settings have default values. If you want to override one or more of the settings for a model, then you must specify the settings with the **params parameter when instantiating the model or later by using the set_params method of the model.
If a parameter is specified by both OML4R algorithm parameters and odm.settings, the value in odm.settings is used.
The following table lists the OML4R parameters that is set by either odm.settings or ore.odm** algorithm explicit parameters.
Table 6-2 OML4R parameters
| Setting Name | OML4R | SQL (DBMS_DATA_MINING) |
|---|---|---|
| ore.odmAI.R | auto.data.prep | PREP_AUTO |
| ore.odmAssocRules.R | case.id.column | CASE_ID_COLUMN_NAME |
| item.id.column | ODMS_ITEM_ID_COLUMN_NAME | |
| item.value.column | ODMS_ITEM_VALUE_COLUMN_NAME | |
| min.support | ASSO_MIN_SUPPORT | |
| min.confidence | ASSO_MIN_CONFIDENCE | |
| max.rule.length | ASSO_MAX_RULE_LENGTH | |
| ore.odmDT.R | auto.data.prep | PREP_AUTO |
| cost.matrix | CLAS_COST_TABLE_NAME | |
| impurity.metric | TREE_IMPURITY_METRIC | |
| max.depth | TREE_TERM_MAX_DEPTH | |
| min.rec.split | TREE_TERM_MINREC_SPLIT | |
| min.pct.split | TREE_TERM_MINPCT_SPLIT | |
| min.rec.node | TREE_TERM_MINREC_NODE | |
| min.pct.node | TREE_TERM_MINPCT_NODE | |
| ore.odmEM.R | num.centers | CLUS_NUM_CLUSTERS |
| auto.data.prep | PREP_AUTO | |
| ore.odmESA.R | auto.data.prep | PREP_AUTO |
| ore.odmESM.R | auto.data.prep | PREP_AUTO |
| ore.odmGLM.R | weights | ODMS_ROW_WEIGHT_COLUMN_NAME |
| type | None in setting, it is mining_function | |
| na.treatment | ODMS_MISSING_VALUE_TREATMENT | |
| reference | GLMS_REFERENCE_CLASS_NAME | |
| ridge | GLMS_RIDGE_REGRESSION | |
| ridge.value | GLMS_RIDGE_VALUE | |
| ridge.vif | GLMS_VIF_FOR_RIDGE | |
| auto.data.prep | PREP_AUTO | |
| ore.odmKMeans.R | auto.data.preps | PREP_AUTO |
| num.centers | CLUS_NUM_CLUSTERS | |
| block.growth | KMNS_BLOCK_GROWTH | |
| conv.tolerance | KMNS_CONV_TOLERANCE | |
| distance.function | KMNS_DISTANCE | |
| iterations | KMNS_ITERATIONS | |
| min.pct.attr.support | KMNS_MIN_PCT_ATTR_SUPPORT | |
| num.bin | KMNS_NUM_BINS | |
| split.criterion | KMNS_SPLIT_CRITERION | |
| ore.odmNB.R | auto.data.prep | PREP_AUTO |
| class.priors | CLAS_PRIORS_TABLE_NAME | |
| ore.odmNMF.R | auto.data.prep | PREP_AUTO |
| num.features | FEAT_NUM_FEATURES | |
| conv.tolerance | NMFS_CONV_TOLERANCE | |
| num.iter | NMFS_NUM_ITERATIONS | |
| rand.seed | NMFS_RANDOM_SEED | |
| allow.negative.scores | NMFS_NONNEGATIVE_SCORING | |
| ore.odmNN.R | type | None in setting, it is mining_function |
| auto.data.prep | PREP_AUTO | |
| ore.odmOC.R | num.centers | CLUS_NUM_CLUSTERS |
| max.buffer | oclt_max_buffer | |
| sensitivity | OCLT_SENSITIVITY | |
| ore.odmRF.R | auto.data.prep | PREP_AUTO |
| ore.odmSVD.R | auto.data.prep | PREP_AUTO |
| ore.odmXGB.R | type | None in setting, it is mining_function |
| auto.data.prep | PREP_AUTO |
For the _init_ method, the argument can be key-value pairs or a
dict. Each list element’s name and value refer to a machine
learning algorithm parameter setting name and value, respectively. The setting value
must be numeric or a string.
The argument for the **params parameter of the set_params method is a dict object mapping a str to a str. The key should be the name of the setting, and the value should be the new setting.
Example 6-2 Specifying Model Settings
This example shows the creation of an Expectation Maximization (EM) model and the changing of a setting.
settings = list(
EMCS_NUM_ITERATIONS= 20,
EMCS_RANDOM_SEED= 7)
EM.MOD <- ore.odmEM(~.-CUST_ID, CUST_DF, num.centers = 3, odm.settings = settings)