R/strategy_exuber.R

In MislavSag/alphar:

library(data.table)
library(exuber)
library(PerformanceAnalytics)
library(ggplot2)
library(runner)
library(future.apply)
library(parallel)
library(fmpcloudr)
library(leanr)
library(mrisk)


# parameters
data_freq = "hour"
data_freq_multiply = 1
use_log = c(TRUE, FALSE) # here starts radf parameters
windows = c(600)
lags = c(0:5)
params <- expand.grid(use_log, windows, lags)
colnames(params) <- c("lag", "window", "lag")

# import data
if (data_freq == "hour") {
  market_data <- import_lean("D:/market_data/equity/usa/hour/trades_adjusted")
  spy <- import_lean("D:/market_data/equity/usa/hour/trades_adjusted")
  risk_path <- file.path("D:/risks/radf-hour")

} else if (data_freq == "minute") {
  market_data <- leanr::get_market_equities_minutes("D:/market_data/equity/usa/minute", "SPY")
  risk_path <- file.path("D:/risks/radf-minute")

  # define sample
  market_data[, time := format.POSIXct(market_data$datetime, "%H:%M:%S")]
  market_data <- market_data[time %between% c("09:30:00", "16:00:00")]

  # upsample
  if (data_freq_multiply == 15) {
    ohlcv <- xts::to.minutes15(as.xts.data.table(market_data[, .(datetime, open, high, low, close, volume)]))
    market_data <- cbind(symbol = market_data$symbol[1], as.data.table(ohlcv))
    setnames(market_data, c("symbol", "datetime", "open", "high", "low", "close", "volume"))
  }
}

# define sample
sp500_stocks <- market_data[, .(symbol, datetime, close)]
sp500_stocks <- sp500_stocks[sp500_stocks[, .N, by = .(symbol)][N > max(windows)], on = "symbol"]

# create directories if necessary
directories <- paste0(risk_path, "/", apply(params, 1, function(x) paste0(x, collapse = "-")))
lapply(directories, function(x) { # create directories if they doesn't exists
  if (!dir.exists(x)) {
    print("Create directory")
    dir.create(x)
  }
})

# choose only empty directories which contain less then 500 files
directories <- directories[lapply(directories, function(x) length(list.files(x))) < 500]

# rolling radf tests
for (dirs in directories) {

  # debuging
  # dirs <- directories[2] # for test
  print(dirs)

  # choose only missing tickers
  symbols <- unique(sp500_stocks$symbol)
  estimated <- gsub('\\.csv', '', list.files(dirs))
  symbols <- setdiff(symbols, estimated)

  # calculate radf values (rolling explosive tests)
  lapply(symbols, function(x) {
    if (!(x %in% c("ACAS", "PLLL", "PETM"))) {
      print(x)
      sample <- sp500_stocks[symbol == x]
      params_ <- stringr::str_split(gsub(".*/", "", dirs), "-")[[1]]
      estimated <- roll_radf(as.data.frame(sample$close),
                             as.logical(as.integer(params_[1])),
                             as.integer(params_[2]),
                             as.integer(params_[3]))
      estimated_appened <- cbind(symbol = x, datetime = sample$datetime, estimated)
      fwrite(estimated_appened, paste0(dirs, "/", x, '.csv'))
    }
  })
}