The SparseSVMPredictor function outputs a table that contains the predicted class of each test sample.
| Column Name | Data Type | Description |
|---|---|---|
| id_column | INTEGER, SMALLINT, BIGINT, NUMERIC, NUMERIC(p), NUMERIC(p,a), TEXT, VARCHAR, VARCHAR(n), UUID, or BYTEA | Contains the identifiers of the test samples. |
| predict_value | VARCHAR | Contains the predicted classes of the test samples. |
| predict_confidence | DOUBLE PRECISION | Contains the prediction confidences. Each prediction confidence is a value between 0 and 1, computed by the formula that follows this table. The higher the value, the more dependable the prediction. |
| accumulate_column | Any | Column copied from the sample table. |
The formula for predict_confidence is:
where i is the attribute id, x i is the value of attributes in the sample, and w i is the weight of attribute i.