Recognising Visual Charting Patterns in the Dow Jones Industrials Index"

In rpatrec: Recognising Visual Charting Patterns in Time Series Data

knitr::opts_chunk$set(collapse = TRUE, comment = "#>")
knitr::opts_chunk$set(fig.width=8, fig.height=5)
library(rpatrec)
library(np)
library(stats)

Applying rpatrec to real-life data

The package comes equipped with a large amount of stock market data for testing purposes. The data is obtained from Datastream. The file \code{stockmarket.RData} contains the daily closing prices of the DAX30, Dow Jones Industrials and the Nikkei 225.

This vignette is not thought to be stand alone documentation but should be read with the report published at the same time.

First, plot the DJI (with patterns marked between colourful lines) and plot the sections that the human observer recognises as patterns. This is our benchmark for this exploration.

dji <- sample.pre(data[[3]])

plot(dji,type="l",main= "Dow Jones Industrials",xlab="Trading Days",ylab="Closing Price",col="blue")
abline(v=400, col="red")
abline(v=1250,col="red")

abline(v=1500, col="green")
abline(v=2100,col="green")

abline(v=2300, col="yellow")
abline(v=2800,col="yellow")

plot(dji[400:1250],type="l",main= "Inverse HS",xlab="Trading Days",ylab="Closing Price",col="blue")
plot(dji[1500:2100],type="l",main= "Dow Jones Industrials",xlab="Trading Days",ylab="Closing Price",col="blue")
plot(dji[2300:2800],type="l",main= "Dow Jones Industrials",xlab="Trading Days",ylab="Closing Price",col="blue")

Now see whether the algorithms can do the same: First, the data is smoothed: We are looking for long-term pattern, so we set the smoother to a high value: Plot again with patterns still highlighted from before

a <- kernel(dji,30)
plot(a,type="l",main= "Dow Jones Industrials",xlab="Trading Days",ylab="Closing Price",col="blue")
abline(v=400, col="red")
abline(v=1250,col="red")

abline(v=1500, col="green")
abline(v=2100,col="green")

abline(v=2300, col="yellow")
abline(v=2800,col="yellow")

Now, run it through the slicer function to see what the package makes of it:

slicer(a,750,150,btpiq=FALSE,rtpiq=FALSE,dtpiq=FALSE)

plot(a[450:1200],type="l",main= "Dow Jones Industrials",xlab="Trading Days",ylab="Closing Price",col="blue")
plot(a[1350:2350],type="l",main= "Dow Jones Industrials",xlab="Trading Days",ylab="Closing Price",col="blue")
plot(a[2100:3300],type="l",main= "Different windows",xlab="Trading Days",ylab="Closing Price",col="blue")

abline(v=0)
abline(v=750)

abline(v=150, col="red")
abline(v=900, col="red")

abline(v=300, col="green")
abline(v=1050, col="green")

abline(v=450,col = "yellow")
abline(v=1200,col = "yellow")

interpret(a[2100:2850])
interpret(a[2250:3000])

For explanation of these results refer to the report.

rpatrec documentation built on May 1, 2019, 11:17 p.m.