Add OneClassSVM model (#97)

* add one_class_svm

* fix shape check

* expand params

Author: sneaxiy

sqlflow_models/__init__.py

@@ -11,6 +11,7 @@
 from .native_keras import RawDNNClassifier
 from .custom_model_example import CustomClassifier
 from .gcn import GCN
+from .one_class_svm import OneClassSVM
 try:
     # NOTE: statsmodels have version conflict on PAI
     from .arima_with_stl_decomposition import ARIMAWithSTLDecomposition

sqlflow_models/one_class_svm.py

@@ -0,0 +1,72 @@
+# Copyright 2020 The SQLFlow Authors. All rights reserved.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+import pickle
+
+import numpy as np
+import tensorflow as tf
+from sklearn.svm import OneClassSVM as SklearnOneClassSVM
+
+MODEL_PATH = "one_class_svm_model"
+
+
+class OneClassSVM(tf.keras.Model):
+    def __init__(self,
+                 feature_columns=None,
+                 kernel='rbf',
+                 degree=3,
+                 gamma='scale',
+                 coef0=0.0,
+                 tol=0.001,
+                 nu=0.5,
+                 shrinking=True,
+                 cache_size=200,
+                 verbose=False,
+                 max_iter=-1):
+        if os.path.exists(MODEL_PATH):
+            with open(MODEL_PATH, "rb") as f:
+                self.svm = pickle.load(f)
+        else:
+            self.svm = SklearnOneClassSVM(kernel=kernel,
+                                          degree=degree,
+                                          gamma=gamma,
+                                          coef0=coef0,
+                                          tol=tol,
+                                          nu=nu,
+                                          shrinking=shrinking,
+                                          cache_size=cache_size,
+                                          verbose=verbose,
+                                          max_iter=max_iter)
+
+    def concat_features(self, features):
+        assert isinstance(features, dict)
+        each_feature = []
+        for _, v in features.items():
+            each_feature.append(v.numpy())
+        return np.concatenate(each_feature, axis=1)
+
+    def sqlflow_train_loop(self, dataset):
+        X = []
+        for features in dataset:
+            X.append(self.concat_features(features))
+        X = np.concatenate(X)
+
+        self.svm.fit(X)
+        with open(MODEL_PATH, "wb") as f:
+            pickle.dump(self.svm, f, protocol=2)
+
+    def sqlflow_predict_one(self, features):
+        features = self.concat_features(features)
+        pred = self.svm.predict(features)
+        return [pred]

tests/test_one_class_svm.py

@@ -0,0 +1,62 @@
+# Copyright 2020 The SQLFlow Authors. All rights reserved.
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import os
+import shutil
+import tempfile
+import unittest
+
+import numpy as np
+import tensorflow as tf
+from sqlflow_models import OneClassSVM
+
+
+class TestOneClassSVM(unittest.TestCase):
+    def setUp(self):
+        self.tmp_dir = tempfile.mkdtemp()
+        self.old_cwd = os.getcwd()
+        os.chdir(self.tmp_dir)
+
+    def tearDown(self):
+        os.chdir(self.old_cwd)
+        shutil.rmtree(self.tmp_dir)
+
+    def create_dataset(self):
+        def generator():
+            for _ in range(10):
+                x1 = np.random.random(size=[1, 1])
+                x2 = np.random.random(size=[1, 1])
+                yield x1, x2
+
+        def dict_mapper(x1, x2):
+            return {"x1": x1, "x2": x2}
+
+        dataset = tf.data.Dataset.from_generator(
+            generator, output_types=(tf.dtypes.float32, tf.dtypes.float32))
+        return dataset.map(dict_mapper)
+
+    def test_main(self):
+        svm = OneClassSVM()
+        train_dataset = self.create_dataset()
+        svm.sqlflow_train_loop(train_dataset)
+
+        predict_dataset = self.create_dataset()
+        for features in predict_dataset:
+            pred = svm.sqlflow_predict_one(features)
+            pred = np.array(pred)
+            self.assertEqual(pred.shape, (1, 1))
+            self.assertTrue(pred[0][0] == 1 or pred[0][0] == -1)
+
+
+if __name__ == '__main__':
+    unittest.main()