The FellegiSunterTrainer function has one output table, which is a model table that is input to the function FellegiSunterPredict. The following table shows the schema of the output table.
Column Name | Data Type | Description |
---|---|---|
_key | VARCHAR | Model property name. |
_value | VARCHAR | Model property value. |
Property Name | Data Type | Description |
---|---|---|
is_supervised | BOOLEAN | Has the value 'true' for supervised learning and 'false' for unsupervised learning. |
comparison_field_cnt | INTEGER | Count of comparison fields, equal to the length of the list specified by the ComparisonFields argument. |
comparison_field_name_i | VARCHAR | Name of comparison field i, where i is in the range [0, comparison_field_cnt-1]. The table has a column for each comparison field. |
comparison_field_threshold_i | DOUBLE PRECISION | Threshold of comparison field i, where i is in the range [0, comparison_field_cnt-1]. If the similarity value exceeds this value, the two objects agree on field i. The table has a column for each comparison field. |
m_i | DOUBLE PRECISION | Probability that the two objects agree on field i, given that the object pair matches, where i is in the range [0, comparison_field_cnt-1]. |
u_i | DOUBLE PRECISION | Probability that the two objects agree on field i, given that the object pair does not match, where i is in the range [0, comparison_field_cnt-1]. |
p | DOUBLE PRECISION | Percentage of object pairs that contain the same object. This column appears only in output for unsupervised learning. |
lower_bound | DOUBLE PRECISION | If the weight of an object pair is less than lower bound, the objects do not match. |
upper_bound | DOUBLE PRECISION | If the weight of an object pair is greater than upper bound, the objects match. |
lambda | DOUBLE PRECISION | Type I (false negative) error, which occurs if an unmatched comparison is erroneously linked. |
mu | DOUBLE PRECISION | Type II (false positive) error, which occurs if a matched comparison is erroneously not linked. |
time_used | DOUBLE PRECISION | Time that the function used to learn the model parameters. |