matman (version 1.1.3)

predictValue: Predicts the value for items

Description

Predicts the value for items based on previous values. Previous values can be aggregated to value per day, week, month, quarter or year. An ARIMA model is estimated for each item based on the function forecast:auto.arima. The best model is selected and used for prediction. Note that only models without drift term will be considered in order to ensure consistent predictions.

Usage

predictValue(
  data,
  value,
  item,
  timestamp,
  temporalAggregation = c("day", "week", "month", "quarter", "year"),
  aggregationFun = sum,
  timeUnitsAhead = 1,
  digits = 3,
  expand = F,
  keepPreviousData = F,
  level = 0.95,
  ...
)

Arguments

data

Data frame including previous values.

value

Name of the column representing the item value.

item

Name of the column representing the item ID or the item name.

timestamp

Name of the column including the timestamp. This column should be in POSIX or date-format.

temporalAggregation

Temporal aggregation mode (i.e., "day", "week", "month", "quarter", "year").

aggregationFun

Function for aggregating the value column. Default is sum.

timeUnitsAhead

Integer indicating the number of time units (i.e., days, weeks, months, quarters or years) the should be predicted.

digits

Integer indicating the number of significant digits used for the predicted values.

expand

Logical indicating whether the data will be expanded after they are aggregated. Default is not (FALSE).

keepPreviousData

Logical indicating whether the data from the given data frame will be added to the result or not. Default is not (FALSE).

level

Numeric value representing the confidence level for the predictions. The default is 0.95 (i.e. lower level = 0.025 and upper level = 0.975).

...

Further arguments for function forecast::auto.arima.

Value

Returns a Forecast object.

See Also

auto.arima Forecast

Examples

# NOT RUN {
# Simple Example
data("Amount")
prediction = predictValue(data = Amount,
    value = "amount",
    item = "item",
    timestamp = "date",
    temporalAggregation = "week",
    timeUnitsAhead = 3)
prediction

# More Sophisticated Example
data("Amount")
prediction = predictValue(data = Amount,
    value = "amount",
    item = "item",
    timestamp = "date",
    temporalAggregation = "week",
    aggregationFun = mean,
    timeUnitsAhead = 5,
    digits = 4,
    keepPreviousData = TRUE,
    level = 0.9,
    trace = TRUE)
prediction
# }