HMMEvaluator accepts four input tables. Three tables contain the HMM parameter tables and output from HMMUnsupervisedLearner or HMMSupervisedLearner function. The fourth table contains the newly observed sequences, and has a schema similar to the input table or views for HMMUnsupervisedLearner.
HMMEvaluator Initial-State Probability Table Schema
| Column |
Data Type |
Description |
|
model_attribute
|
ANY |
The name of the column that contains the model attribute, specified by the InitStateModelColumn attribute. |
|
state_key_attribute
|
ANY |
A symbol defining a hidden state. |
| probability |
DOUBLE PRECISION |
The initial probability of the hidden state. |
HMMEvaluator State-Transition Probability Table Schema
| Column |
Data Type |
Description |
|
model_attribute
|
ANY |
The name of the column that contains the model attribute, specified by the TransAttributeColumn attribute. |
|
from_state_key_attribute
|
ANY |
A symbol defining a hidden state from which a transition emanates. |
|
to_state_key_attribute
|
ANY |
A symbol defining a hidden state to which a transition is made. |
| probability |
DOUBLE PRECISION |
The probability of a transition from the from_state to the to_state. |
HMMEvaluator Emission Probability Table Schema
| Column |
Data Type |
Description |
|
model_attribute
|
ANY |
The name of the column that contains the model attribute, specified by the EmitModelColumn attribute. |
|
state_key_attribute
|
ANY |
A symbol defining a hidden state. |
|
observed_key_attribute
|
ANY |
A symbol defining an observation value. |
| probability |
DOUBLE PRECISION |
The probability of emitting an observed symbol given that the HMM is in a hidden state. |