Einführung in die Vorverarbeitung
Vorverarbeitung für Machine Learning in Python
James Chapman
Curriculum Manager, DataCamp
Was ist Datenvorverarbeitung?
- Nach explorativer Datenanalyse und Datenbereinigung
- Daten fürs Modellieren vorbereiten
- Beispiel: kategoriale Merkmale in numerische umwandeln (Dummy-Variablen)
Warum vorverarbeiten?
- Datensatz so transformieren, dass er zum Modellieren passt
- Modellleistung verbessern
- Zuverlässigere Ergebnisse erzeugen
Rückblick: Daten mit pandas erkunden
import pandas as pd
hiking = pd.read_json("hiking.json")
print(hiking.head())
Prop_ID Name ... lat lon
0 B057 Salt Marsh Nature Trail ... NaN NaN
1 B073 Lullwater ... NaN NaN
2 B073 Midwood ... NaN NaN
3 B073 Peninsula ... NaN NaN
4 B073 Waterfall ... NaN NaN
Rückblick: Daten mit pandas erkunden
print(hiking.info())
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 33 entries, 0 to 32
Data columns (total 11 columns):
# Column Non-Null Count Dtype
-- ------ -------------- -----
0 Prop_ID 33 non-null object
1 Name 33 non-null object
2 Location 33 non-null object
3 Park_Name 33 non-null object
4 Length 29 non-null object
5 Difficulty 27 non-null object
6 Other_Details 31 non-null object
7 Accessible 33 non-null object
8 Limited_Access 33 non-null object
9 lat 0 non-null float64
10 lon 0 non-null float64
dtypes: float64(2), object(9)
memory usage: 3.0+ KB
Rückblick: Daten mit pandas erkunden
print(wine.describe())
Type Alcohol ... Alcalinity of ash
count 178.000000 178.000000 ... 178.000000
mean 1.938202 13.000618 ... 19.494944
std 0.775035 0.811827 ... 3.339564
min 1.000000 11.030000 ... 10.600000
25% 1.000000 12.362500 ... 17.200000
50% 2.000000 13.050000 ... 19.500000
75% 3.000000 13.677500 ... 21.500000
max 3.000000 14.830000 ... 30.000000
Fehlende Daten entfernen
print(df)
A B C
0 1.0 NaN 2.0
1 4.0 7.0 3.0
2 7.0 NaN NaN
3 NaN 7.0 NaN
4 5.0 9.0 7.0
print(df.dropna())
A B C
1 4.0 7.0 3.0
4 5.0 9.0 7.0
Fehlende Daten entfernen
print(df)
A B C
0 1.0 NaN 2.0
1 4.0 7.0 3.0
2 7.0 NaN NaN
3 NaN 7.0 NaN
4 5.0 9.0 7.0
print(df.drop([1, 2, 3]))
A B C
0 1.0 NaN 2.0
4 5.0 9.0 7.0
Fehlende Daten entfernen
print(df)
A B C
0 1.0 NaN 2.0
1 4.0 7.0 3.0
2 7.0 NaN NaN
3 NaN 7.0 NaN
4 5.0 9.0 7.0
print(df.drop("A", axis=1))
B C
0 NaN 2.0
1 7.0 3.0
2 NaN NaN
3 7.0 NaN
4 9.0 7.0
Fehlende Daten entfernen
print(df)
A B C
0 1.0 NaN 2.0
1 4.0 7.0 3.0
2 7.0 NaN NaN
3 NaN 7.0 NaN
4 5.0 9.0 7.0
print(df.isna().sum())
A 1
B 2
C 2
dtype: int64
print(df.dropna(subset=["B"]))
A B C
1 4.0 7.0 3.0
3 NaN 7.0 NaN
4 5.0 9.0 7.0
Fehlende Daten entfernen
print(df)
A B C
0 1.0 NaN 2.0
1 4.0 7.0 3.0
2 7.0 NaN NaN
3 NaN 7.0 NaN
4 5.0 9.0 7.0
print(df.dropna(thresh=2))
A B C
0 1.0 NaN 2.0
1 4.0 7.0 3.0
4 5.0 9.0 7.0
Lass uns üben!
Vorverarbeitung für Machine Learning in Python
