FcGradientBoosting: Forecast Gradient Boosting

In Mthrun/TSAT: Time Series Analysis Tools (TSAT)

Forecast Gradient Boosting

Description

Gradient Boosting as published by [Chen/Guestrin, 2016] is a machine learning technique for regression and classification problems, which produces a prediction model in the form of the ensemble of weak prediction models, typically decision trees. However it is claimed that an ensemble of weak predictors when combined together should hive a strong model with superior accuracy.

Usage

FcGradientBoosting(DataVec, SplitAt, Time, ForecastHorizon, Frequency = "days", FUN = sum, PlotIt = TRUE)

Arguments

DataVec	[1:n] numerical vector of time series data
SplitAt	Index of row where the DataVec is divided into test and train data. If not given n is used
Time	[1:n] character vector of Time in the length of data
ForecastHorizon	Scalar defining the timesteps to forecast ahead
Frequency	If aggregation required. Can be either "days", "weeks", "months" or "quarters" or "years", see ConvertNumerical2TSobject. Default is "days"
FUN	Aggregation function, usually sum
PlotIt	FALSE (default), do nothing. TRUE: Plots the forecast (versus the test set)

Details

"Gradient Boosting algorithm is a machine learning technique used for building predictive tree-based models.Boosting is an ensemble technique in which new models are added to correct the errors made by existing models. Models are added sequentially until no further improvements can be made. The ensemble technique uses the tree ensemble model which is a set of classification and regression trees (CART)." [see url]

Do not confuse with bagging: simple ensembling technique in which we build many independent predictors/models/learners and combine them using some model averaging techniques. (e.g. weighted average, majority vote or normal average.

It should be noted that in order to forecast more than one step saisonality extracted by either an arma model (gbm) or and fourier model (forecastxgb) has to be used.

Value

List with

Forecast	vector
Model	Model object of class xgbar, see xgbar
ForecastStats	Forecast statistics object of class forecast, see forecast package
TestData	[1:n-SplitAt, 1:2] Data frame of test data and test time
TrainData	[1:SplitAt, 1:2] Data frame of training data and training time

Note

forecastxgb not on cran, available through: devtools::install_github("ellisp/forecastxgb-r-package/pkg")

Author(s)

Michael Thrun

References

[Chen/Guestrin, 2016] Chen, Tianqi ans Guestrin, Carlos: XGBoost: A Scalable Tree Boosting System, In Krishnapuram, Balaji Shah, Mohak Smola, Alexander J. Aggarwal, Charu C., Shen, Dou, Rastogi, Rajeev, Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, San Francisco, CA, USA, August 13-17, 2016, ACM, pp. 785–794. arXiv:1603.02754, doi:10.1145/2939672.2939785, 2016.

https://medium.com/@rakesh.poojary/forecasting-markets-using-extreme-gradient-boosting-xgboost-9c62605c4b83

See Also

xgbar

Examples

data('ElectricityBRD')
# Forecast on a test data set containing the last 30 values
SplitAt = length(ElectricityBRD$Mrd_KWh)-30
res = FcGradientBoosting(ElectricityBRD$Mrd_KWh, SplitAt = SplitAt,Time = ElectricityBRD$Time , Frequency='months', ForecastHorizon=30, PlotIt=TRUE)

Mthrun/TSAT documentation built on Feb. 5, 2024, 11:15 p.m.