New Multivariate Drift Method: Classifier for Drift Detection (#355)

* adding cdd * flake8 and mypy fixes * expose calculator * filter categorical feature warnings * CDD draft doc * add cdd docs * doc update * py37 compatibility update * update docs * doc update * doc update * Break up the big chunk calculation block * Formatting --------- Co-authored-by: Niels Nuyttens <niels@nannyml.com>
NannyML · Feb 16, 2024 · 3e11cd7 · 3e11cd7
1 parent 808913c
commit 3e11cd7
Show file tree

Hide file tree

Showing 31 changed files with 13,710 additions and 11,976 deletions.
diff --git a/.gitignore b/.gitignore
@@ -133,3 +133,6 @@ config/*
 
 # generated by docker build
 requirements.txt
+
+# dev debug folder
+debug/
diff --git a/docs/_static/butterfly-multivariate-drift.svg b/docs/_static/butterfly-multivariate-drift.svg
diff --git a/docs/_static/butterfly-scatterplot.svg b/docs/_static/butterfly-scatterplot.svg
diff --git a/docs/_static/how-it-works/butterfly-multivariate-drift-cdd.svg b/docs/_static/how-it-works/butterfly-multivariate-drift-cdd.svg
diff --git a/docs/_static/how-it-works/butterfly-multivariate-drift-pca.svg b/docs/_static/how-it-works/butterfly-multivariate-drift-pca.svg
diff --git a/docs/_static/how-it-works/butterfly-scatterplot.svg b/docs/_static/how-it-works/butterfly-scatterplot.svg
diff --git a/...terfly-univariate-drift-distributions.svg → ...terfly-univariate-drift-distributions.svg b/...terfly-univariate-drift-distributions.svg → ...terfly-univariate-drift-distributions.svg
diff --git a/...ting_data_drift/multivariate_drift_detection/classifier-for-drift-detection.svg b/...ting_data_drift/multivariate_drift_detection/classifier-for-drift-detection.svg
diff --git a/.../detecting_data_drift/multivariate_drift_detection/pca-reconstruction-error.svg b/.../detecting_data_drift/multivariate_drift_detection/pca-reconstruction-error.svg
diff --git a/...orks - Data Reconstruction with PCA.ipynb → ...s/How It Works - Multivariate Drift.ipynb b/...orks - Data Reconstruction with PCA.ipynb → ...s/How It Works - Multivariate Drift.ipynb
@@ -109,7 +109,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "figure.savefig(\"../_static/butterfly-scatterplot.svg\")"
+    "figure.savefig(\"../_static/how-it-works/butterfly-scatterplot.svg\")"
    ]
   },
   {
@@ -153,7 +153,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "figure.write_image('../_static/butterfly-univariate-drift-distributions.svg')"
+    "figure.write_image('../_static/how-it-works/butterfly-univariate-drift-distributions.svg')"
    ]
   },
   {
@@ -173,8 +173,8 @@
     "rcerror_results = rcerror_calculator.calculate(data=analysis_df)\n",
     "\n",
     "# let's visualize results across all the dataset\n",
-    "figure = rcerror_results.plot()\n",
-    "figure.show()"
+    "figure1 = rcerror_results.plot()\n",
+    "figure1.show()"
    ]
   },
   {
@@ -184,7 +184,38 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "figure.write_image('../_static/butterfly-multivariate-drift.svg')"
+    "figure1.write_image('../_static/how-it-works/butterfly-multivariate-drift-pca.svg')"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "2a9d81bd-90dd-42b2-a16e-39847e250041",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# Let's compute multivariate drift\n",
+    "drift_classifier = nml.DriftDetectionClassifierCalculator(\n",
+    "    feature_column_names=feature_column_names,\n",
+    "    timestamp_column_name='ordered',\n",
+    "    chunk_size=DPP\n",
+    ").fit(reference_data=reference_df)\n",
+    "# let's compute results for analysis period\n",
+    "drift_classifier_results = drift_classifier.calculate(data=analysis_df)\n",
+    "\n",
+    "# let's visualize results across all the dataset\n",
+    "figure2 = drift_classifier_results.plot()\n",
+    "figure2.show()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "dece03d8-69c3-4a54-893e-80aae8d0897d",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "figure2.write_image('../_static/how-it-works/butterfly-multivariate-drift-cdd.svg')"
    ]
   }
  ],
@@ -204,7 +235,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.8.8"
+   "version": "3.11.7"
   }
  },
  "nbformat": 4,