KNNRecommender Arguments - Teradata Vantage

Machine Learning Engine Analytic Function Reference

Product
Teradata Vantage
Release Number
8.00
1.0
Published
May 2019
Language
English (United States)
Last Update
2019-11-22
dita:mapPath
blj1506016597986.ditamap
dita:ditavalPath
blj1506016597986.ditaval
dita:id
B700-4003
lifecycle
previous
Product Category
Teradata Vantageā„¢
WeightModelTable
Specify the name for the output table of interpolation weights.
BiasModelTable
Specify the name for the output table of global, user, and item bias statistics.
NearestItemsTable
[Optional] Specify the name for the output table of nearest neighbors for each item.

If you omit this argument, the function does not output this table.

If you specify this argument, and a table named item_neighbors_table exists, the function uses the existing table to train the model.

If you specify this argument, and no table named item_neighbors_table exists, the function creates a table with that name.

UserIDColumn
[Optional] Specify the user_rating_table column that contains the user IDs.
Default: First user_rating_table column
You must specify either all or none of UserIDColumn, ItemIDColumn, and RatingColumn.
ItemIDColumn
[Optional] Specify the user_rating_table column that contains the item IDs.
Default: Second user_rating_table column
RatingColumn
[Optional] Specify the user_rating_table column that contains the ratings.
Default: Third user_rating_table column
K
[Optional] Specify the number of nearest neighbors with which to calculate the interpolation weights.
Default: 20
LearningRate
[Optional] Specify the initial learning rate. The learning rate adjusts automatically during training based on changes in the root-mean-square error (RMSE).
Default: 0.001
MaxIterNum
[Optional] Specify the maximum number of iterations.
Default: 10
StopThreshold
[Optional] Specify the RMSE below which the function stops.
Default: 0.0002
SimilarityMethod
[Optional] Specify the method for calculating item similarity:
Option Description
'pearson' (Default) Pearson correlation coefficient:


'adjustedcosine' Adjusted cosine similarity: