Merge pull request #191 from beckernick/confusion-matrix-cleanup

switch cupy_conf_mat to confusion_matrix

gpu_bdb/bdb_tools/cupy_metrics.py

@@ -17,32 +17,6 @@
 import cupy as cp
 
 
-def cupy_conf_mat(y, y_pred):
-    """
-    Simple, fast confusion matrix for two class models designed to match sklearn.
-    Assumes the classes are one of [0, 1]. It will fail edge cases, which are fairly
-    numerous.
-
-    Implementation taken from rapidsai/cuml#1524
-    """
-    nclasses = len(cp.unique(y))
-    assert nclasses == 2
-    res = cp.zeros((2, 2))
-
-    pos_pred_ix = cp.where(y_pred == 1)
-    neg_pred_ix = cp.where(y_pred != 1)
-    tn_sum = (y[neg_pred_ix] == 0).sum()
-    fn_sum = (y[neg_pred_ix] == 1).sum()
-    tp_sum = (y[pos_pred_ix] == 1).sum()
-    fp_sum = (y[pos_pred_ix] == 0).sum()
-
-    res[0, 0] = tn_sum
-    res[1, 0] = fn_sum
-    res[0, 1] = fp_sum
-    res[1, 1] = tp_sum
-    return res
-
-
 def cupy_precision_score(y, y_pred):
     """
     Simple precision score method for two class models.

gpu_bdb/queries/q05/gpu_bdb_query_05.py

@@ -25,7 +25,7 @@
 )
 
 from bdb_tools.readers import build_reader
-from bdb_tools.cupy_metrics import cupy_conf_mat, cupy_precision_score
+from bdb_tools.cupy_metrics import cupy_precision_score
 
 import cupy as cp
 import numpy as np
@@ -78,6 +78,7 @@ def build_and_predict_model(ml_input_df):
     Returns the model and accuracy statistics
     """
     import cuml
+    from cuml.metrics import confusion_matrix
 
     feature_names = ["college_education", "male"] + [
         "clicks_in_%d" % i for i in range(1, 8)
@@ -105,7 +106,9 @@ def build_and_predict_model(ml_input_df):
 
     results_dict["auc"] = roc_auc_score(y.to_array(), y_pred.to_array())
     results_dict["precision"] = cupy_precision_score(cp.asarray(y), cp.asarray(y_pred))
-    results_dict["confusion_matrix"] = cupy_conf_mat(cp.asarray(y), cp.asarray(y_pred))
+    results_dict["confusion_matrix"] = confusion_matrix(
+        cp.asarray(y, dtype="int32"), cp.asarray(y_pred, dtype="int32")
+    )
     results_dict["output_type"] = "supervised"
     return results_dict
 

gpu_bdb/queries/q05/gpu_bdb_query_05_sql.py

@@ -29,7 +29,7 @@
     gpubdb_argparser,
     run_query,
 )
-from bdb_tools.cupy_metrics import cupy_conf_mat, cupy_precision_score
+from bdb_tools.cupy_metrics import cupy_precision_score
 from sklearn.metrics import roc_auc_score
 import cupy as cp
 
@@ -58,6 +58,7 @@ def build_and_predict_model(ml_input_df):
     Returns the model and accuracy statistics
     """
     import cuml
+    from cuml.metrics import confusion_matrix
 
     feature_names = ["college_education", "male"] + [
         "clicks_in_%d" % i for i in range(1, 8)
@@ -85,7 +86,9 @@ def build_and_predict_model(ml_input_df):
 
     results_dict["auc"] = roc_auc_score(y.to_array(), y_pred.to_array())
     results_dict["precision"] = cupy_precision_score(cp.asarray(y), cp.asarray(y_pred))
-    results_dict["confusion_matrix"] = cupy_conf_mat(cp.asarray(y), cp.asarray(y_pred))
+    results_dict["confusion_matrix"] = confusion_matrix(
+        cp.asarray(y, dtype="int32"), cp.asarray(y_pred, dtype="int32")
+    )
     results_dict["output_type"] = "supervised"
     return results_dict