In mowomoyela/tsfcmethodr: Classification methods to predict best performing forecasting method of a ts

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  cache = TRUE,
  message = FALSE,
  warning = FALSE
)

tsfcmethodr

Provides four different classification machine learning methods and one C5.0 rule model to predict the best performing forecasting method for time series. The four different classification methods are: xgboost, cat boost, svm and ann. This package contains labeled time series data to train the models for the prediction of new time series data.

Usage

# Initialize the R package
library(tsfcmethodr)
# Example of a xgboost classification model with the basic ts 
# taxonomy from tstaxonomyr R package --------
  # Train a xgboost model 
  fitted_model <- train_xgb(n_round = 10, cv_nfold = 10, 
                            tune_length = 10, ts_taxonomy = "v1")
  # Predict best performing forecasting method for a new classified ts
  ts_sales = datasets::BJsales
  prediction <- predict_fc_model(fitted_model, ts_sales, "v1")
  prediction

# Example of a svm classification model with the ligther feature 
# selected ts taxonomy from tstaxonomyr R package --------
  # Train a svm model 
  fitted_model <- train_svm(n_round = 10, cv_nfold = 10,
                            tune_length = 10, ts_taxonomy = "v2")
  # Predict best performing forecasting method for a new classified ts
  ts_sales = datasets::BJsales
  prediction <- predict_fc_model(fitted_model, ts_sales, "v2")
  prediction

Installation

You can install the development version 1.0.0 from Github with:

devtools::install_github("mowomoyela/tsfcmethdr")

Overview

All provided functions of this package:

• Version 1.0.0
  • train_svm: Generates a classification svm trainings model.
  • train_ann: Generates a classification ann trainings model.
  • train_xgb: Generates a classification xgboost trainings model.
  • train_c50_rule_model: Generates a C5.0 rule model.
  • train_catboost: Generates a classification catboost trainings model.
  • predict_fc_model: Predicts the best performing forecasting method for a new time series.

License

This package is free and open source software, licensed under GPL-2.

mowomoyela/tsfcmethodr documentation built on May 15, 2019, 4:47 p.m.