In antononcube/QRMon-R: Quantile Regression Monad

library(QRMon)
library(SparseMatrixRecommender)
library(quantreg)
library(magrittr)
library(splines)
library(ggplot2)
library(ExternalParsersHookUp)

Introduction

This notebook demonstrates the construction of Quantile Regression workflows using natural language commands.

Basic pipeline

Simple creation:

qrmon <-
  QRMonUnit( dfTemperatureData )

pipeline <- 
  ToQuantileRegressionWorkflowCode( 
    "use object qrmon;
     compute quantile regression with 12 knots;
     show plot", parse = T )
pipeline

res <- 
  eval( expr = ToQuantileRegressionWorkflowCode( 
    "use object qrmon;
     summarize data;
     show plot" ) )

 ToQuantileRegressionWorkflowCode( 
    "create from dfTemperatureData;
     compute quantile regression with 12 knots;
     show plot", parse = F )

qrObj2 <-
  eval( expr = ToQuantileRegressionWorkflowCode( 
    "create from dfTemperatureData;
     summarize data;
     rescale value axis;
     compute quantile regression with 12 knots;
     show plot" ) )

Outliers

ToQuantileRegressionWorkflowCode( 
    "use object qrmon;
     compute quantile regression with knots 12, interpolation degree 2 and probabilities 0.1 0.5 0.95;
     show date list plot with date origin 1900-01-01;
     find outliers", parse = F)

qrmon2 <- 
  eval( expr = ToQuantileRegressionWorkflowCode( 
    "use object qrmon;
     compute quantile regression with knots 12 and probabilities 0.1 0.5 0.95;
     show date list plot with date origin 1900-01-01;
     find outliers", parse = T) )

Anomalies

ToQuantileRegressionWorkflowCode( 
    "create from dfTemperatureData;
     compute quantile regression with 16 knots and probability 0.5;
     show date list plot with date origin 1900-01-01;
     show absolute errors plot;
     find anomalies by the threshold 5;
     take pipeline value;
    ", parse=F)

dfAnomalies <-
  eval( expr = ToQuantileRegressionWorkflowCode( 
    "create from dfTemperatureData;
     compute quantile regression with 16 knots and probability 0.5;
     show date list plot with date origin 1900-01-01;
     show absolute errors plot;
     echo text anomalies finding follows;
     find anomalies by the threshold 5;
     take pipeline value;
    ") )

Plot the data (dfTemperatureData) and the found anomalies:

(QRMonUnit(dfTemperatureData) %>% QRMonPlot(echoQ = FALSE, datePlotQ = TRUE, dateOrigin = "1900-01-01") %>% QRMonTakeValue) + 
ggplot2::geom_point( data = dfAnomalies, ggplot2::aes( x = as.POSIXct(Regressor, origin = "1900-01-01"), y = Value ), color = "red" )