R/portfolio_optimizer.R

In PortfolioAnalysis: Portfolio Optimization Methods

#' @title Optimize portfolio weights
#' @description Optimizes portfolio weights by minimizing the variance
#' @param asset.names Vector of ticker of securities
#' @param increment Number of portfolio to be generated, Default: 100
#' @param rf Risk-free rate of return, Default: 0
#' @param period Period for which the returns are calculated, Default: c("months", "weeks", "quarters", "years")
#' @return Returns a dataframe of the portfolios with different portfolio weights and different target returns and target standard deviation. Using this an investor can choose between range of portfolio to allocate funds.
#' @details Minimizes the variance using the method of lagrange multiplier to calculate the portfolio weights with minimized variance for given target return.
#' @examples
#' optim.portfolio(c('FXAIX', 'TIBFX'), period = 'days')
#' @rdname optim.portfolio
#' @import quantmod
#' @importFrom purrr map
#' @import PerformanceAnalytics
#' @importFrom stats cov
#' @importFrom xts to.period merge.xts
#' @import quadprog
#' @importFrom lubridate years
#' @export
optim.portfolio = function(asset.names, increment = 100, rf = 0, period = c('months', 'weeks', 'quarters', 'years')){
dat = map(.x = asset.names, .f = function(x){
  raw = getSymbols(x, auto.assign = FALSE)
  raw = to.period(raw, period = period)
})
dat_clean = map(.x = 1:length(asset.names), function(x){
  trim = paste0(format(Sys.Date() - years(5), '%Y'), '/')
  dat = dat[[x]][trim,6]
  dat = CalculateReturns(dat)
})
Ret.monthly = do.call(merge.xts, dat_clean)
Ret.monthly = Ret.monthly[-1,]
colnames(Ret.monthly) = asset.names

mat.ret = matrix(Ret.monthly, nrow = nrow(Ret.monthly))
colnames(mat.ret) = c(asset.names)
VCOV = cov(mat.ret)
avg.ret = matrix(apply(mat.ret, 2, mean), nrow = ncol(Ret.monthly), ncol = 1)
rownames(avg.ret) = c(asset.names)
colnames(avg.ret) = c('avg.ret')
min.ret = min(avg.ret)
max.ret = max(avg.ret)
increment = increment
tgt.ret = seq(min.ret, max.ret, length.out = increment)
tgt.sd = rep(0, length(tgt.ret))
wgt = matrix(0, nrow = increment, ncol = length(avg.ret))

Dmat = 2*VCOV
dvec = avg.ret
Amat = cbind(1, avg.ret, diag(nrow(avg.ret)))
for(i in 1:100){
bvec = c(1, tgt.ret[i], rep(0, nrow(avg.ret)))
soln = solve.QP(Dmat = Dmat, dvec = dvec, Amat = Amat, bvec = bvec, meq = 2)
wgt[i,] = soln$solution
wgt = ifelse(wgt<0.00001, 0, wgt)
tgt.sd[i] = sqrt(wgt[i,] %*% VCOV %*% wgt[i,])
}
colnames(wgt) = c(paste0(asset.names, rep('.wgt', nrow(avg.ret))))
sharpe_ratio = (tgt.ret - rf)/tgt.sd
port_summary = cbind(tgt.ret,tgt.sd, wgt, sharpe_ratio)
port_summary = as.data.frame(port_summary)
port_summary
}