When the function completes, KNNRecommenderTrain displays a table of the root mean square error (rmse) at each iteration (schema shown in the following table).
Column Name | Data Type | Description |
---|---|---|
iternum | Integer | Iteration number. The baseline rmse shows null in this column. |
rmse | Double | Root mean square error after that iteration. |
The function also creates the following three output tables: a table of the interpolation weights, a table of the bias values calculated by the function, and an optional table of nearest (item) neighbors.
Column Name | Data Type | Description |
---|---|---|
itemid | Integer | Item id. |
weights | bytea | Interpolation weights for that item in compressed binary format. |
Column Name | Data Type | Description |
---|---|---|
label | char(1) | U - user statistics I - item statistics G - global statistics |
id | Integer | Item id or user id. |
value | Double | If label = I, average rating across all users for that item. If label = U, average rating across all items from that user. If label = G, global average rating across all users and all items. |
Column Name | Data Type | Description |
---|---|---|
itemi | Integer | Item id of item i. |
itemj | Integer | Item id of item j. |
sij | Double | Calculated similarity between item i and item j. |