look at dataset balance

3springs · Oct 21, 2020 · efbb0ce · efbb0ce
1 parent 5eedb2b
commit efbb0ce
Show file tree

Hide file tree

Showing 2 changed files with 95 additions and 20 deletions.
diff --git a/notebooks/b05_Supervised_Learning/supervised_part1.ipynb b/notebooks/b05_Supervised_Learning/supervised_part1.ipynb
diff --git a/notebooks/b05_Supervised_Learning/supervised_part1.py b/notebooks/b05_Supervised_Learning/supervised_part1.py
@@ -149,6 +149,17 @@
 y = geolink["LITHOLOGY_GEOLINK"]
 X
 
+# NOTE: our dataset is imbalanced, this is always import when considering a problem.
+#     
+# 1. so our baseline accuracy is 46.5
+# 2. if we get poor performance may want to consider techniques to deal with unbalanced data. However we do not do this in the notebook
+
+# Check dataset label balance
+counts = y.value_counts()
+counts = counts[counts>0]/counts.sum()
+counts.plot.bar()
+counts
+
 X_train, X_test, y_train, y_test = train_test_split(
     X, y, test_size=0.2, random_state=2020
 )