Fixbacktest (PaddlePaddle#153)

* fix backtest * fix backtest
AmberC0209 · Oct 24, 2022 · b9a9826 · b9a9826
1 parent 68a7f5f
commit b9a9826
Show file tree

Hide file tree

Showing 3 changed files with 12 additions and 11 deletions.
diff --git a/docs/locale/zh_CN/LC_MESSAGES/source/modules/backtest/overview.po b/docs/locale/zh_CN/LC_MESSAGES/source/modules/backtest/overview.po
@@ -162,7 +162,7 @@ msgid ""
 " set to simulate the realy prediction."
 msgstr ""
 "``predict_window`` 是每次预测的窗口长度。``stride`` "
-"是两次连续预测之间的移动步长。在大多数情况下我们需要自定义这两个参数来模拟真是的预测场景。"
+"是两次连续预测之间的移动步长。在大多数情况下我们需要自定义这两个参数来模拟真实的预测场景。"
 
 #: ../../source/modules/backtest/overview.rst:123
 #: e4d054512d7440d3a79f7444fae9f856

diff --git a/paddlets/tests/utils/test_backtest.py b/paddlets/tests/utils/test_backtest.py
@@ -42,8 +42,8 @@ def setUp(self):
             ))
         known_cov = TimeSeries.load_from_dataframe(
             pd.DataFrame(
-                np.random.randn(2500, 2).astype(np.float32),
-                index=pd.date_range("2022-01-01", periods=2500, freq="15T"),
+                np.random.randn(2000, 2).astype(np.float32),
+                index=pd.date_range("2022-01-01", periods=2000, freq="15T"),
                 columns=["b1", "c1"]
             ))
         static_cov = {"f": 1, "g": 2}
@@ -83,9 +83,8 @@ def test_backtest(self):
         score, predicts = backtest(self.tsdataset1, lstnet, start=pd.Timestamp('2022-01-07T12'), predict_window=50, stride=50,
                        return_predicts=True)
 
-        start = 624
-        data_len = len(self.tsdataset1.get_target())
-        assert len(predicts.get_target()) == data_len - start
+
+        assert len(predicts.get_target()) == 1300
 
         # case3 add window,stride, window != stride
         lstnet = LSTNetRegressor(
@@ -107,9 +106,8 @@ def test_backtest(self):
         lstnet.fit(self.tsdataset1, self.tsdataset1)
         score, predicts = backtest(self.tsdataset1, lstnet, start=200, predict_window=50, stride=50, return_predicts=True)
 
-        start = 200 + 4 * 4
-        data_len = len(self.tsdataset1.get_target())
-        assert len(predicts.get_target()) == data_len - start
+
+        assert len(predicts.get_target()) == 1700
 
         # case5 add return score
         lstnet = LSTNetRegressor(

diff --git a/paddlets/utils/backtest.py b/paddlets/utils/backtest.py
@@ -118,12 +118,15 @@ def _check():
     for _ in tqdm(range(predict_rounds), desc=TQDM_PREFIX, disable=not verbose):
         data._target, rest = all_target.split(index) 
         rest_len = len(rest)
-        if rest_len < predict_window + model_skip_chunk_len:
+
+        if rest_len < model_out_chunk_len + model_skip_chunk_len:
             if data.known_cov is not None:
                 target_end_time = data._target.end_time
                 known_index = data.known_cov.get_index_at_point(target_end_time)
-                if len(data.known_cov) - known_index - 1 < predict_window + model_skip_chunk_len:
+                if len(data.known_cov) - known_index - 1 < model_out_chunk_len + model_skip_chunk_len:
                     break
+
+        if rest_len < predict_window + model_skip_chunk_len:
             predict_window = rest_len
             predict_window = predict_window - model_skip_chunk_len