데이터 정제
Python으로 IoT 데이터 분석하기
Matthias Voppichler
IT Developer
결측치
IoT 기기에서 결측치가 생기는 이유
- 불안정한 네트워크 연결
- 전원 부족
- 기타 외부 요인
데이터 품질을 다루는 시점
- 수집 중
- 분석 중
결측치 처리
결측치 처리 방법
- 채우기
- 평균
- 중앙값
- 이전값으로 채우기(Forward-fill)
- 다음값으로 채우기(Backward-fill)
- 삭제
- 분석 중단
결측치 탐지
df.info()
<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 12 entries, 2018-10-15 08:00:00 to 2018-10-15 08:55:00
Data columns (total 3 columns):
temperature 8 non-null float64
humidity 8 non-null float64
precipitation 12 non-null float64
dtypes: float64(3)
memory usage: 384.0 bytes
결측치 삭제
print(df.head())
temperature humidity precipitation
timestamp
2018-10-15 08:00:00 16.7 64.2 0.0
2018-10-15 08:05:00 16.6 NaN 0.0
2018-10-15 08:10:00 16.5 65.3 0.0
2018-10-15 08:15:00 NaN 65.0 0.0
2018-10-15 08:20:00 16.8 64.3 0.0
df.dropna()
temperature humidity precipitation
timestamp
2018-10-15 08:00:00 16.7 64.2 0.0
2018-10-15 08:10:00 16.5 65.3 0.0
2018-10-15 08:20:00 16.8 64.3 0.0
결측치 채우기
df
temperature humidity precipitation
timestamp
2018-10-15 08:00:00 16.7 64.2 0.0
2018-10-15 08:05:00 16.6 NaN 0.0
2018-10-15 08:10:00 17.0 65.3 0.0
2018-10-15 08:15:00 NaN 65.0 0.0
2018-10-15 08:20:00 16.8 64.3 0.0
df.fillna(method="ffill")
temperature humidity precipitation
timestamp
2018-10-15 08:00:00 16.7 64.2 0.0
2018-10-15 08:05:00 16.6 64.2 0.0
2018-10-15 08:10:00 17.0 65.3 0.0
2018-10-15 08:15:00 17.0 65.0 0.0
2018-10-15 08:20:00 16.8 64.3 0.0
중단된 측정
print(df.head())
timestamp temperature humidity
2018-10-15 00:00:00 13.5 84.7
2018-10-15 00:10:00 13.3 85.6
2018-10-15 00:20:00 12.9 88.8
2018-10-15 00:30:00 12.8 89.2
2018-10-15 00:40:00 13.0 87.7
print(df.isna().sum())
temperature 0
humidity 0
dtype: int64
df_res = df.resample("10min").last()
print(df_res.head())
timestamp temperature humidity
2018-10-15 00:00:00 13.5 84.7
2018-10-15 00:10:00 13.3 85.6
2018-10-15 00:20:00 12.9 88.8
2018-10-15 00:30:00 12.8 89.2
2018-10-15 00:40:00 13.0 87.7
print(df_res.isna().sum())
temperature 34
humidity 34
dtype: int64
중단된 측정
df_res.plot(title="Environment")
연습해 봅시다!
Python으로 IoT 데이터 분석하기
