The unsupervised model is generated by specifying the user defined threshold values for the different metrics in the ComparisonFields argument. The initialization parameters InitialP, InitialM, InitialU, Lambda, and Mu are set to their default values. The column match_tag is not used for unsupervised learning.
DROP TABLE IF EXISTS fg_unsupervised_model; CREATE DIMENSION TABLE "fg_unsupervised_model" AS SELECT * FROM FellegiSunterTrainer ( ON (SELECT 1) PARTITION BY 1 InputTable ('fstrainer_input ') ComparisonFields ('jaro1_sim: 0.8', 'ld1_sim:0.8', 'ngram1_sim:0.5', 'jw1_sim:0.8') InitialP (0.1) InitialM (0.9) InitialU (0.1) Lambda (0.9) Mu (0.9) );