transformação por PCA

pyFTS/common/transformations/pca.py

@@ -0,0 +1,36 @@
+from sklearn.decomposition import PCA
+from pyFTS.common.Transformations import Transformation
+import pandas as pd
+
+class PCATransformation(Transformation):
+    def __init__(self):
+        self.pca = PCA(n_components=2)
+        self.is_multivariate = True
+
+
+    def apply(self, data, param=None, **kwargs):
+        endogen_variable = kwargs.get('endogen_variable', None)
+        names = kwargs.get('names', ('x', 'y'))
+        if endogen_variable not in data.columns:
+            endogen_variable = None
+        cols = data.columns[:-1] if endogen_variable is None else [col for col in data.columns if
+                                                                   col != endogen_variable]
+        self.pca.fit(data[cols].values)
+        transformed = self.pca.transform(data[cols])
+        new = pd.DataFrame(transformed, columns=list(names))
+        new[endogen_variable] = data[endogen_variable].values
+        return new
+
+if __name__ =="__main__":
+    pd.set_option('max_columns', 50)
+
+    file = '/home/matheus_cascalho/Documentos/matheus_cascalho/MINDS/TimeSeries_Lab/SOM/gas_concentration/ethylene_CO.csv'
+
+    df = pd.read_csv(file)
+    df = df[df['Time (seconds)'].apply(lambda x: x % 1 == 0)]
+    ignore = list(df.columns)[:3]
+    endogen_variable = 'TGS2602'
+    pca = PCATransformation()
+    cols = [col for col in df.columns if col not in ignore]
+    # cols.append(endogen_variable)
+    print(pca.apply(df[cols], endogen_variable=endogen_variable))