Create Lag and Moving Average Features

Add lagged and moving average features for target and exogenous variables. This function creates features suitable for time series modeling by computing lags and rolling averages within groups.

Usage

feat_lag_ma_dt(
  df,
  date,
  target,
  p = NULL,
  q = NULL,
  groups = NULL,
  xreg = NULL,
  xreg_lags = NULL,
  xreg_ma = NULL
)

Arguments

df

A data frame containing time series data

date

Character string naming the date column

target

Character string naming the target variable column

p

Integer, number of target lags to create (e.g., p=12 creates lag_1 through lag_12)

q

Integer vector, moving average window sizes for target (e.g., q=c(7,28))

groups

Character vector naming grouping columns (e.g., c("store", "item"))

xreg

Character vector naming exogenous variable columns to transform

xreg_lags

Named list of lag specifications for exogenous variables (e.g., list(price = c(1,7), promotion = c(0,1)))

xreg_ma

Named list of MA window specifications for exogenous variables (e.g., list(price = c(7,28)))

Value

Data frame with original columns plus new lag and MA features

Details

Lags and MAs are computed within each group separately. The data is automatically sorted by groups and date before computation. Use lag 0 in xreg_lags to include the current value of an exogenous variable.

Examples

if (FALSE) { # \dontrun{
# Create 7 lags and 7-day MA for sales
df_feat <- feat_lag_ma_dt(df, date = "date", target = "sales",
                           p = 7, q = 7, groups = "store_id")

# Add exogenous variable features
df_feat <- feat_lag_ma_dt(df, date = "date", target = "sales",
                           p = 3, groups = "store_id",
                           xreg = "price",
                           xreg_lags = list(price = c(0, 1, 7)))
} # }