Getting Started with chronofeat • chronofeat

Overview

chronofeat is an R package for time series feature engineering and forecasting. It provides:

Formula-based feature specification: Define lags, moving averages, rolling statistics, and calendar features using a concise formula syntax
Model-agnostic design: Works with any R model that has a fit/predict interface (lm, glm, xgboost, randomForest, etc.)
Recursive multi-step forecasting: Automatically generates features at each forecast step
Panel data support: Handle multiple time series (groups) in a single workflow

Installation

# Install from GitHub
remotes::install_github("quantics/chronofeat")

library(chronofeat)
library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union

Quick Start

The typical chronofeat workflow has three steps:

Prepare data with TimeSeries() (optional but recommended)
Fit a model with fit() using a feature formula
Generate forecasts with forecast()

Step 1: Load and Prepare Data

chronofeat includes a sample retail sales dataset with 42 product categories (items) and monthly observations:

data(retail)
head(retail)
#> # A tibble: 6 × 3
#>   date       items value
#>   <date>     <fct> <dbl>
#> 1 1982-04-01 V10    94  
#> 2 1982-05-01 V10   106. 
#> 3 1982-06-01 V10    95.1
#> 4 1982-07-01 V10    95.3
#> 5 1982-08-01 V10    82.8
#> 6 1982-09-01 V10    89.4

Create a TimeSeries object to bundle the data with frequency information:

ts_data <- TimeSeries(
  data = retail,
  date = "date",
  groups = "items",
  frequency = "month"
)

ts_data
#> TimeSeries object
#> ----------------
#> Date column: date ( Date )
#> Frequency: month 
#> Groups: items 
#> Observations: 13986 
#> 
#> Data (first few rows):
#> # A tibble: 6 × 3
#>   date       items value
#>   <date>     <fct> <dbl>
#> 1 1982-04-01 V10    94  
#> 2 1982-05-01 V10   106. 
#> 3 1982-06-01 V10    95.1
#> 4 1982-07-01 V10    95.3
#> 5 1982-08-01 V10    82.8
#> 6 1982-09-01 V10    89.4

The TimeSeries object:

Automatically sorts data by groups and date
Detects or validates the time frequency
Provides optional preprocessing (gap-filling, calendar completion)

Step 2: Fit a Model

Use fit() with a formula to specify your target variable and features:

model <- fit(
  value ~ p(12) + q(7, 12) + month(),
  data = ts_data,
  model = lm
)

This formula creates:

p(12): 12 lags of the target (value_lag_1 through value_lag_12)
q(7, 12): 7-period and 12-period moving averages
month(): Month as a factor variable

The model object contains the fitted model and all metadata needed for forecasting:

# Check what predictors were created
model$predictors
#>  [1] "value_lag_1"  "value_lag_2"  "value_lag_3"  "value_lag_4"  "value_lag_5" 
#>  [6] "value_lag_6"  "value_lag_7"  "value_lag_8"  "value_lag_9"  "value_lag_10"
#> [11] "value_lag_11" "value_lag_12" "value_ma_7"   "value_ma_12"  "month"

Step 3: Generate Forecasts

Use forecast() to generate multi-step ahead predictions:

fc <- forecast(model, h = 6)
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
#> Warning in predict.lm(object, newdata = newdata, ...): prediction from
#> rank-deficient fit; attr(*, "non-estim") has doubtful cases
head(fc)
#> # A tibble: 6 × 3
#>   items date       value_forecast
#>   <fct> <date>              <dbl>
#> 1 V10   2010-01-01           380.
#> 2 V10   2010-02-01           357.
#> 3 V10   2010-03-01           344.
#> 4 V10   2010-04-01           364.
#> 5 V10   2010-05-01           378.
#> 6 V10   2010-06-01           409.

The forecast is recursive: each predicted value becomes history for generating features at the next step.

Visualizing Results

library(ggplot2)

# Get one item for plotting
item_a <- retail %>%
  filter(items == levels(items)[1]) %>%
  tail(36)

fc_a <- fc %>%
  filter(items == levels(retail$items)[1])

ggplot() +
  geom_line(data = item_a, aes(x = date, y = value), color = "black") +
  geom_line(data = fc_a, aes(x = date, y = value_forecast), color = "blue") +
  geom_point(data = fc_a, aes(x = date, y = value_forecast), color = "blue") +
  labs(title = "Actual vs Forecast", x = "Date", y = "Value") +
  theme_minimal()

Understanding the Formula Syntax

The formula’s left side specifies the target variable. The right side specifies features:

Syntax	Description	Example Output
`p(k)`	k lags of target	`value_lag_1`, …, `value_lag_k`
`p(1, 7, 12)`	Specific lags	`value_lag_1`, `value_lag_7`, `value_lag_12`
`q(w1, w2)`	Moving averages	`value_ma_w1`, `value_ma_w2`
`dow()`	Day of week	`dow` (factor)
`month()`	Month	`month` (factor)
`woy()`	Week of year	`woy` (integer)
`eom()`	End of month	`eom` (0/1)
`dom()`	Day of month	`dom` (integer)
`rollsum(w)`	Rolling sum	`value_rollsum_w`
`rollsd(w)`	Rolling std dev	`value_rollsd_w`
`rollmin(w)`	Rolling minimum	`value_rollmin_w`
`rollmax(w)`	Rolling maximum	`value_rollmax_w`
`rollslope(w)`	Rolling trend	`value_rollslope_w`
`trend(d)`	Polynomial trend	`trend1`, `trend2`, …
`lag(var, k)`	Lag of exogenous	`var_lag_k`
`ma(var, w)`	MA of exogenous	`var_ma_w`
`varname`	Raw column	Include as-is

Example Formulas

# Basic autoregressive model
value ~ p(12)

# Add calendar seasonality
value ~ p(12) + month() + dow()

# Add rolling statistics
value ~ p(12) + q(7) + rollsum(7, 28) + rollsd(7)

# Include exogenous variable with lags
value ~ p(12) + lag(price, 0, 1, 7) + promo

# Complex model
value ~ p(1, 7, 12) + q(7, 28) + month() + dow() + rollslope(12) + trend(1)

Using Different Models

chronofeat works with any model that has a standard R interface. Pass the model function directly:

# Linear regression
m_lm <- fit(value ~ p(12) + month(), data = ts_data, model = lm)

# GLM for count data
m_glm <- fit(count ~ p(7) + dow(), data = count_data,
             date = "date", model = glm, family = poisson())

# Random Forest
library(randomForest)
m_rf <- fit(value ~ p(12) + month(), data = ts_data, model = randomForest)

For more control, provide a custom model specification with fit and predict functions:

xgb_spec <- list(
  fit = function(y, X, ...) {
    xgboost::xgboost(
      data = as.matrix(X),
      label = y,
      nrounds = 100,
      verbose = 0,
      ...
    )
  },
  predict = function(object, newdata, ...) {
    predict(object, as.matrix(newdata))
  }
)

m_xgb <- fit(value ~ p(12) + month(), data = ts_data, model = xgb_spec)

See the Building Custom Models article for detailed examples with XGBoost, LightGBM, and other frameworks.

Panel Data (Multiple Time Series)

chronofeat handles panel data (multiple time series) through the groups parameter. Each group gets its own:

Lag and MA calculations (no cross-contamination)
Forecasts (predictions stay within group boundaries)

# The retail dataset has 42 items
n_distinct(retail$items)
#> [1] 42

# fit() handles all groups automatically
model <- fit(
  value ~ p(12) + month(),
  data = ts_data,
  model = lm
)

# forecast() produces forecasts for each group
fc <- forecast(model, h = 3)
fc %>%
  group_by(items) %>%
  summarise(n = n()) %>%
  head()
#> # A tibble: 6 × 2
#>   items     n
#>   <fct> <int>
#> 1 V10       3
#> 2 V11       3
#> 3 V12       3
#> 4 V13       3
#> 5 V14       3
#> 6 V15       3

Cross-Validation

Evaluate model performance with time series cross-validation:

cv_results <- cv_forecast(
  value ~ p(12) + month(),
  data = ts_data,
  model = lm,
  h = 6,
  n_windows = 5,
  metric = "rmse"
)

print(cv_results)

See the Cross-Validation article for more details.

Next Steps

Building Custom Models: Learn how to integrate XGBoost, LightGBM, neural networks, and custom ensembles
Feature Engineering Reference: Complete guide to all available features
Data Preprocessing: Handle missing dates, gaps, and irregular calendars
Cross-Validation: Proper evaluation workflows for time series
Advanced Workflows: Hyperparameter tuning, model comparison, and production patterns

Key Functions Reference

Function	Purpose
`TimeSeries()`	Create a time series object with preprocessing
`fit()`	Fit a forecasting model with feature formula
`forecast()`	Generate recursive multi-step forecasts
`cv_forecast()`	Time series cross-validation
`as_model_spec()`	Convert model function to specification