Lag মানে previous time period-এর value।
Daily sales:
Date Sales
D1 100
D2 120
D3 90
Lag 1:
Date Sales Lag1
D1 100 NaN
D2 120 100
D3 90 120
Pandas:
df["sales_lag_1"] = df["sales"].shift(1)
Multiple Products থাকলে
df = df.sort_values(
["product_id", "date"]
)
df["lag_1"] = (
df.groupby("product_id")["sales"]
.shift(1)
)
Very Important
আগে sorting ঠিক করতে হবে।
না হলে lag conceptually wrong হয়ে যাবে।
Forecasting-এ এটা common bug।