在得到各位成员的帮助后，我终于在quantstrat中建立了自己的样本策略。该代码在＆lt; 100种股票的世界中运行良好且快速（约1分钟），但是当宇宙扩展到例如500种股票时，运行时间急剧增加到（> 5小时）。经过一些测试后，我发现增加运行时间的关键是applyStrategy。重新平衡。当我消除重新平衡这个词时，一切都很快。
但是，我希望保持再平衡选项以维持一定的股权与现金比率。任何关于大大增加运行时间和可能的解决方案的想法将不胜感激。
为了解释更多，我的策略相当简单。每个月，投资组合将出售投资组合中的所有股票，并回购排名前20位的股票。资金平均分配给20只股票。
具有更大Universe的数据文件： （请编辑代码中的路径）
http://www.filedropper.com/data7
http://www.filedropper.com/data72
我的代码:(在小样本下运行良好）
library(lattice)
library(quantstrat)
quotes.file <- "Data7.csv" #Need Editing
quotes.file2 <- "Data72.csv" #Need Editing
#
# 1. Instrument: read csv file into xts object
#
code <- read.csv(quotes.file2,head=TRUE)
data2 <- read.csv(quotes.file, head=TRUE)
code=apply(code,1,"toString")
matsplitter<-function(M, r, c) {
rg <- (row(M)-1)%/%r+1
cg <- (col(M)-1)%/%c+1
rci <- (rg-1)*max(cg) + cg
N <- prod(dim(M))/r/c
cv <- unlist(lapply(1:N, function(x) M[rci==x]))
dim(cv)<-c(r,c,N)
cv
}
data=matsplitter(data2[,-1],nrow(data2),4)
colnames(data)=c("close","pe","pe_avg","pe_sd")
dates <- as.Date(data2$Date,"%d/%m/%Y")
C=list()
i=1
for (i in 1:(dim(data)[3])){
C[[i]] <- xts(x=data[,,i], order.by=dates)
}
#
# 2. Support Functions
#
osFixedDollar <- function(timestamp,orderqty, portfolio, symbol, ruletype, ...)
{
updatePortf(qs.portfolio, Symbols=code, Dates=paste('::',as.Date(Sys.time()),sep=''))
cumPL=0
for (o in (1: length(code))){
cumPL <- cumPL+sum(getPortfolio(portfolio)$symbols[[code[o]]]$posPL$Net.Trading.PL)
}
ClosePrice <- as.numeric((mktdata[timestamp,"close"]))
orderqty <- round(((initEq+cumPL)/topchoice)/ClosePrice,-2)
return(orderqty)
}
pause <- function() {
cat ("Press [enter] to continue")
line <- readline()
}
Zscore <- function(x0,y0,z0) {
x <- as.numeric(x0)
y <- as.numeric(y0)
z <- as.numeric(z0)
res <- (x-y)/z
res
}
for (i in 1:(dim(data)[3])){
C[[i]]$Zscore <-Zscore(C[[i]]$pe,C[[i]]$pe_avg,C[[i]]$pe_sd)
}
for (j in 1:(nrow(data))){
temp=0
for (i in 1:(dim(data)[3])){
temp=c(temp,C[[i]][j,"Zscore"])
}
temp=temp[-1]
temp=as.numeric(-temp)
temp2=rank(temp)
for (i in 1:(dim(data)[3])){
C[[i]]$ranking[j]=temp2[i]
}
}
for (i in 1:(dim(data)[3])){
assign(code[i],C[[i]])
}
#
# 3. Initializations
#
# Currency:
currency("HKD")
# Instrument:
big.point.value <- 1
for (i in 1:(dim(data)[3])){
stock(code[i], currency="HKD", multiplier=big.point.value)
}
# Strategy, portfolio and account names:
qs.strategy <- "PE.on.HKex"
qs.portfolio <- "Equities"
qs.account <- "My.account"
# Remove previous strategy, portfolio and account:
suppressWarnings(rm(list = c(paste("account", qs.account, sep='.'), paste("portfolio", qs.portfolio, sep='.')), pos=.blotter))
suppressWarnings(rm(list = c(qs.strategy, paste("order_book", qs.portfolio, sep='.')), pos=.strategy))
rm.strat(qs.strategy) # remove strategy etc. if this is a re-run
# Portfolio:
initDate <- '2013-12-15'
initPortf(qs.portfolio, code, initDate=initDate, currency='HKD')
# Account:
initEq <- 100000
initAcct(qs.account, portfolios=qs.portfolio, initDate=initDate, currency='HKD', initEq=initEq)
# Orders:
initOrders(portfolio=qs.portfolio, initDate=initDate)
# Strategy:
strategy(qs.strategy, store=TRUE)
#
# 4. Functions indicating buy, sell short, sell and buy to cover
#
nb.bars.with.no.trade <- 0
topchoice=20
buy.function <- function(x) {
N <- nrow(x)
buy.signals <- numeric(N)
# buy.signals[1:nb.bars.with.no.trade]<- 0
for (i in (nb.bars.with.no.trade+1):N) {
if ( (x[i]<=topchoice)
&& (index(x)[i] < index(get(code[1]))[nrow(data)]) && (index(x)[i] > index(get(code[1]))[1]) # trick to avoid an open trade at the end
) {
buy.signals[i] <- 1
} else {
buy.signals[i] <- 0
}
}
xts(x=buy.signals, order.by=index(x))
}
sell.function <- function(x) {
N <- nrow(x)
sell.signals <- numeric(N)
#sell.signals[1:nb.bars.with.no.trade]<- 0
for (i in (nb.bars.with.no.trade+1):N) {
if (x[i]>topchoice){
sell.signals[i] <- 1
}
}
xts(x=sell.signals, order.by=index(x))
}
#
# 5. Indicators for buy, sell short, sell and buy to cover
#
add.indicator(strategy = qs.strategy,
name = "buy.function",
arguments = list(x = quote(mktdata[,"ranking"])),
label="Buy")
add.indicator(strategy = qs.strategy,
name = "sell.function",
arguments = list(x = quote(mktdata[,"ranking"])),
label="Sell")
#
# 6. Signals for buy, sell short, sell and buy to cover
#
add.signal(qs.strategy,
name="sigFormula",
arguments = list(formula="X1.Buy == 1"),
label="Buy",
cross=FALSE)
add.signal(qs.strategy,
name="sigFormula",
arguments = list(formula="X1.Sell == 1"),
label="Sell",
cross=FALSE)
#
# 7. Rules for buy, sell short, sell and buy to cover
#
nb.contracts <- 2
round.trip.commission.for.one.contract <- 6
round.trip.commission.for.total.contracts <- round.trip.commission.for.one.contract * nb.contracts
add.rule(qs.strategy,
name='ruleSignal',
arguments = list(sigcol="Sell",
sigval=TRUE,
orderqty='all',
ordertype='market',
orderside='long',
pricemethod='market',
replace=FALSE,TxnFees=-0),
type='exit',
label="LongExit")
add.rule(qs.strategy,
name='ruleSignal',
arguments = list(sigcol="Buy",
sigval=TRUE,
orderqty=100000000,
ordertype='market',
orderside='long',osFUN='osMaxPos',
replace=FALSE),
type='enter',
label="LongEntry")
add.rule(qs.strategy, 'rulePctEquity',
arguments=list(rebalance_on='years',
trade.percent=(1/topchoice)*0.9,
refprice=quote(last(getPrice(mktdata)[paste('::',curIndex,sep='')])),
digits=0
),
type='rebalance',
label='rebalance'
)
posval<-initEq/topchoice
for(symbol in code){
pos<-round((posval/first(get(symbol)[,1])[,1]),0)
addPosLimit(qs.portfolio,symbol,initDate, maxpos=pos,minpos=-pos)
}
#
# 8. Apply Strategy to Porfolio
#
tradeSize <- initEq/topchoice
applyStrategy.rebalancing(strategy = qs.strategy,
portfolios = qs.portfolio)
updatePortf(qs.portfolio, Symbols=code, Dates=paste('::',as.Date(Sys.time()),sep=''))
updateAcct(qs.account)
updateEndEq(qs.account)