The HMMUnsupervised function outputs a message and outputs three tables through JDBC.
Output Message Schema
Column | Data Type | Description |
---|---|---|
message | VARCHAR | Reports that the models will be saved to the output tables after the training process successfully completes. |
InitStateTable Schema
The function saves this table to init_state_prob, whose default value is Pi in the current schema.
Column | Data Type | Description |
---|---|---|
model_column | BIGINT, INTEGER, or VARCHAR | [Column appears only with ModelColumn syntax element.] Model attribute. |
state | VARCHAR | Hidden state of learned HMM. |
probability | DOUBLE PRECISION | Initial state probability determined by function. |
StateTransitionTable Schema
The function saves this table to state_transition_prob, whose default value is A in the current schema.
Column | Data Type | Description |
---|---|---|
model_column | BIGINT, INTEGER, or VARCHAR | [Column appears only with ModelColumn syntax element.] Model attribute. |
from_state | VARCHAR | Hidden state of learned HMM, from which a transition emanates. |
to_state | VARCHAR | Hidden state of learned HMM, to which a transition is made. |
probability | DOUBLE PRECISION | Probability of a transition from from_state to to_state. |
EmissionTable Schema
The function saves this table to emit_prob, whose default value is B in the current schema.
Column | Data Type | Description |
---|---|---|
model_column | BIGINT, INTEGER, or VARCHAR | [Column appears only with ModelColumn syntax element.] Model attribute. |
state | VARCHAR | Hidden state of learned HMM. |
observed | VARCHAR | Symbol in observed_id column of input table. |
probability | DOUBLE PRECISION | Probability of emitting observed symbol, given that HMM is in hidden state. |