com.github.jelmerk.spark.knn.evaluation
Param for the column name for the approximate results.
Param for the column name for the approximate results. Default: "approximateNeighbors"
Returns the accuracy of the approximate results.
Returns the accuracy of the approximate results.
a dataset
the accuracy of the approximate results
Param for the column name for the exact results.
Param for the column name for the exact results. Default: "exactNeighbors"
identifier
identifier
Evaluator for knn algorithms, which expects two input columns, the exact neighbors and approximate neighbors. It compares the results to determine the accuracy of the approximate results. Typically you will want to compute this over a small sample given the cost of computing the exact results on a large index.