Lære Engineering Multiple Lags

Stryg for at vise menuen

Selecting the right lag intervals is crucial when you want to capture longer-term dependencies in time series data. Lag variables help your model learn from previous observations, but deciding which lags to include is a balancing act. Shorter lags, such as lag-1, can capture immediate effects, while longer lags like lag-3 or lag-7 might reveal weekly or other cyclical patterns. The choice depends on the frequency and seasonality of your data, as well as the specific patterns you expect to find. However, adding more lag features increases the dimensionality of your dataset, which can lead to overfitting and higher computational cost. You should always consider the trade-off between capturing more information and keeping your model simple and generalizable.


              123456789101112131415
            
import pandas as pd

# Example time series DataFrame
data = {
    "date": pd.date_range(start="2024-01-01", periods=10, freq="D"),
    "value": [10, 12, 13, 15, 14, 16, 17, 19, 18, 20]
}
df = pd.DataFrame(data)

# Generate multiple lag features: lag-1, lag-3, lag-7
df["lag_1"] = df["value"].shift(1)
df["lag_3"] = df["value"].shift(3)
df["lag_7"] = df["value"].shift(7)

print(df)

Var alt klart?

Tak for dine kommentarer!

Sektion 1. Kapitel 3

Spørg AI

Spørg om hvad som helst eller prøv et af de foreslåede spørgsmål for at starte vores chat

Sektion 1. Kapitel 3