Preparar e visualizar dados incrementais

Analisando Dados de IoT em Python

Matthias Voppichler

IT Developer

Preparação de dados

Fazer pivot
Reamostrar
Aplicar diff()
Aplicar pct_change()

Estrutura dos dados

print(data.head())

                timestamp device   value
0 2018-10-26 06:30:42.817   C331  6020.0
1 2018-10-26 06:30:43.083   M640   104.0
2 2018-10-26 06:31:00.858   M640   126.0
3 2018-10-26 06:31:10.254   C331  6068.0
4 2018-10-26 06:31:10.474   M640   136.0

Tabela dinâmica (pivot)

Gráfico explicando pivot-table

Aplicar tabela dinâmica

                timestamp device   value
0 2018-10-26 06:30:42.817   C331  6020.0
1 2018-10-26 06:30:43.083   M640   104.0
2 2018-10-26 06:31:00.858   M640   126.0
3 2018-10-26 06:31:10.254   C331  6068.0
4 2018-10-26 06:31:10.474   M640   136.0

data = pd.pivot_table(data, columns="device", values="value", index="timestamp")
print(data.head()

device                     C331   M640
timestamp                             
2018-10-26 06:30:42.817  6020.0    NaN
2018-10-26 06:30:43.083     NaN  104.0
2018-10-26 06:31:00.858     NaN  126.0
2018-10-26 06:31:10.254  6068.0    NaN
2018-10-26 06:31:10.474     NaN  136.0

Reamostrar

# Reamostrar DataFrame para 1min
df = data.resample("1min").max().dropna()
print(df.head())

device                 C331   M640
timestamp                         
2018-10-26 06:30:00  6020.0  104.0
2018-10-26 06:31:00  6129.0  180.0
2018-10-26 06:32:00  6205.0  256.0
2018-10-26 06:33:00  6336.0  332.0
2018-10-26 06:34:00  6431.0  402.0

Visualizar dados

data.plot()
plt.show()

pd.diff()

# Diferença
df_diff = data.diff(1)
df_diff.plot()
plt.show()

Análise de dados — diferença

# Diferença
df_diff = data.diff()
df_diff.plot()
plt.show()

# Diferença reamostrada
df = data.resample('30min').max()
df_diff = df.diff()
df_diff.plot()
plt.show()

Variação percentual

df_pct = df_diff.pct_change()
df_pct.plot()

Vamos praticar!

Analisando Dados de IoT em Python