Merge pull request #76 from MarcCoru/master

Fixed parsing of UEA UCR datasets

tslearn/datasets.py

@@ -58,6 +58,27 @@ def extract_from_zip_url(url, target_dir=None, verbose=False):
             sys.stderr.write("Corrupted zip file encountered, aborting.\n")
         return None
 
+def in_file_string_replace(filename, old_string, new_string):
+    """ String replacement within a text file. It is used to fix typos in downloaded csv file.
+    The code was modified from "https://stackoverflow.com/questions/4128144/replace-string-within-file-contents"
+
+    Parameters
+    ----------
+    filename : string
+        Path to the file where strings should be replaced
+    old_string : str
+        The string to be replaced in the file.
+    new_string : str
+        The new string that will replace old_string
+
+    """
+    with open(filename) as f:
+        s = f.read()
+
+    with open(filename, 'w') as f:
+        s = s.replace(old_string, new_string)
+        f.write(s)
+
 
 class UCR_UEA_datasets(object):
     """A convenience class to access UCR/UEA time series datasets.
@@ -89,29 +110,14 @@ def __init__(self, use_cache=True):
             url_baseline = "http://www.timeseriesclassification.com/singleTrainTest.csv"
             self._baseline_scores_filename = os.path.join(self._data_dir, os.path.basename(url_baseline))
             urlretrieve(url_baseline, self._baseline_scores_filename)
+
+            # fix typos in that CSV to match with the name in the download link
+            in_file_string_replace(self._baseline_scores_filename, "CinCECGtorso", "CinCECGTorso")
+            in_file_string_replace(self._baseline_scores_filename, "StarlightCurves", "StarLightCurves")
         except:
             self._baseline_scores_filename = None
 
         self._ignore_list = ["Data Descriptions"]
-        # File names for datasets for which it is not obvious
-        self._filenames = {"CinCECGtorso": "CinC_ECG_torso",
-                           "CricketX": "Cricket_X",
-                           "CricketY": "Cricket_Y",
-                           "CricketZ": "Cricket_Z",
-                           "FiftyWords": "50words",
-                           "Lightning2": "Lighting2",
-                           "Lightning7": "Lighting7",
-                           "NonInvasiveFatalECGThorax1": "NonInvasiveFetalECG_Thorax1",
-                           "NonInvasiveFatalECGThorax2": "NonInvasiveFetalECG_Thorax2",
-                           "GunPoint": "Gun_Point",
-                           "SonyAIBORobotSurface1": "SonyAIBORobotSurface",
-                           "SonyAIBORobotSurface2": "SonyAIBORobotSurfaceII",
-                           "SyntheticControl": "synthetic_control",
-                           "TwoPatterns": "Two_Patterns",
-                           "UWaveGestureLibraryX": "UWaveGestureLibrary_X",
-                           "UWaveGestureLibraryY": "UWaveGestureLibrary_Y",
-                           "UWaveGestureLibraryZ": "UWaveGestureLibrary_Z",
-                           "WordSynonyms": "WordsSynonyms"}
 
     def baseline_accuracy(self, list_datasets=None, list_methods=None):
         """Report baseline performances as provided by UEA/UCR website.
@@ -209,23 +215,29 @@ def load_dataset(self, dataset_name):
         (1000, 128, 1)
         >>> print(y_train.shape)
         (1000,)
+        >>> X_train, y_train, X_test, y_test = UCR_UEA_datasets().load_dataset("StarLightCurves")
+        >>> print(X_train.shape)
+        (1000, 1024, 1)
+        >>> X_train, y_train, X_test, y_test = UCR_UEA_datasets().load_dataset("CinCECGTorso")
+        >>> print(X_train.shape)
+        (40, 1639, 1)
         >>> X_train, y_train, X_test, y_test = UCR_UEA_datasets().load_dataset("DatasetThatDoesNotExist")
         >>> print(X_train)
         None
         """
         full_path = os.path.join(self._data_dir, dataset_name)
-        fname_train = self._filenames.get(dataset_name, dataset_name) + "_TRAIN.txt"
-        fname_test = self._filenames.get(dataset_name, dataset_name) + "_TEST.txt"
+        fname_train = dataset_name + "_TRAIN.txt"
+        fname_test = dataset_name + "_TEST.txt"
         if not os.path.exists(os.path.join(full_path, fname_train)) or \
             not os.path.exists(os.path.join(full_path, fname_test)):
             url = "http://www.timeseriesclassification.com/Downloads/%s.zip" % dataset_name
             for fname in [fname_train, fname_test]:
                 if os.path.exists(os.path.join(full_path, fname)):
                     os.remove(os.path.join(full_path, fname))
-            extract_from_zip_url(url, target_dir=self._data_dir, verbose=False)
+            extract_from_zip_url(url, target_dir=full_path, verbose=False)
         try:
-            data_train = numpy.loadtxt(os.path.join(full_path, fname_train), delimiter=",")
-            data_test = numpy.loadtxt(os.path.join(full_path, fname_test), delimiter=",")
+            data_train = numpy.loadtxt(os.path.join(full_path, fname_train), delimiter=None)
+            data_test = numpy.loadtxt(os.path.join(full_path, fname_test), delimiter=None)
         except:
             return None, None, None, None
         X_train = to_time_series_dataset(data_train[:, 1:])