Anaconda3 on Ubuntu 18.04
|
1 2 3 4 5 6 7 |
cd~ wget https://repo.continuum.io/archive/Anaconda3-5.0.1-Linux-x86_64.sh bash Anaconda3-5.0.1-Linux-x86_64.sh -b -p ~/anaconda rm Anaconda3-5.0.1-Linux-x86_64.sh echo 'export PATH="~/anaconda/bin:$PATH"' >> ~/.bashrc source .bashrc conda update conda |
Attempted to install forecast package in R 3.2 version, on Ubuntu 16.04 LTS Xenial Xerus:
|
1 |
install.packages("forecast") |
but when installing dependent packages tseries, quadprod, RcppArmadillo, it produced an error:
/usr/bin/ld: cannot find -lblas, cannot find -llalapack
Solution: install these libs in linux terminal:
|
1 |
sudo apt install libblas-dev liblapack-dev |
and install.packages("forecast") will work in R console
Today we will use R astsa (Applied Statistical Time Series Analysis) library to analyze data and predict it with ARIMA model.
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 |
library(curl) library(quantmod) library(rusquant) library(xts) library(utils) # ASTSA lib is for ARIMA library(astsa) # Load the data price.ohlc <- getSymbols("SBER", from = Sys.Date()-1, to = Sys.Date(), src = "Finam", period = "1min", auto.assign = FALSE) price.cl <- Cl(price.ohlc) # Display close price on plot plot(price.cl) |
Compare auto correlation and partial autocorrelation functions for diff data
|
1 |
asf2(na.omit(diff(price.cl)) |
Output: Continue reading
R code template to predict the future with ARIMA auto regression. To try different models, change the line with fit <- arima(...)
|
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 |
# Rusquant is quantmod library with ability # to load Russian market with intraday data # Use vanilla quantmod for other markets library(rusquant) # Load symbol from data provider data.raw <- getSymbols("SBER", src = "Finam", period="5min", from = "2017-08-25", to = "2017-08-25", auto.assign = FALSE) # Convert price to stationary data data <- diff(log(Cl(data.raw))) # Train auto regression arima model fit <- arima(data, order=c(10,0,1)) # Predict the future pred <- predict(fit, n.ahead=100) # Xts index lost during prediction, so convert predicted data to xts pred.index <- as.POSIXct(index(pred$pred), origin = min(index(data))) pred.xts <- xts(pred$pred, order.by = pred.index) # Draw original and predicted values on the same chart plot(c(data, pred.xts), type = "n") lines(data) lines(pred.xts, col = "red") |
