R/features-basic.R
In ConvergenceDA/visdom: R package for energy data analytics
Defines functions
basicFeatures
Documented in
basicFeatures
# Copyright 2016 The Board of Trustees of the Leland Stanford Junior University.
# Direct inquiries to Sam Borgeson (sborgeson@stanford.edu)
# or professor Ram Rajagopal (ramr@stanford.edu)
#' @title
#' Implementation of several basic statistical features
#'
#' @description
#' Function that implements a set of common statistical features that can be run on an instance of \code{\link{MeterDataClass}}.
#'
#' @param meterData An instance of \code{\link{MeterDataClass}} that provides the data used for all the implemented feature calculations.
#'
#' @details
#' \code{basicFeatures} is called by passing in an instance of a MeterDataClass, which provides the meter data in both linear series
#' and day-per-row matrix formats. These data structures and some associated indices and intermediate data structures are
#' re-used throughout this method, which tries to take advantage of the performance benefits of such re-use.
#' This function is often called in the context of the \code{iterator.*} suite of functions, which can loop through large sets of
#' meters, calling feature functions on each.
#'
#' @examples
#' \dontrun{
#' DATA_SOURCE = TestData(n=10)
#' meterData = DATA_SOURCE$getMeterDataClass(DATA_SOURCE$getIds()[1])
#' features = basicFeatures(meterData)
#' names(features)
#' class(features)
#' }
#' @export
basicFeatures = function(meterData, ...){ # r is an instance of MeterDataClass
data <- as.matrix(meterData$kwMat)
id <- meterData$id
monthTotals = sapply(split(meterData$kw,factor(format(meterData$dates,'%b'),levels=month.abb)),mean,na.rm=T) * 30 * 24
names(monthTotals) <- paste('kw.total.',names(monthTotals),sep='')
monthTotals = as.list(monthTotals) # needs to be a list ofr concatenation to work
summerMon = 4:8 # May through Sept - zero based
summerSubset = as.POSIXlt(meterData$dates)$mon %in% summerMon
wkdySubset = as.POSIXlt(meterData$dates)$day %in% 1:5
# todo: apply/quantile is slow. Is it worth softening min,max this way?
qtle <- apply(data,1,FUN=quantile,c(0.01,0.99),na.rm=T)
dMin <- qtle[1,]
dMax <- qtle[2,]
hMean <- colMeans(data,na.rm=T)
names(hMean) <- paste('HOD.mean.',1:24,sep='')
dMean <- rowMeans(data,dims=1,na.rm=T) # rowMeans is faster than apply with FUN=mean ?rowMeans for details
tMean <- rowMeans(as.matrix(meterData$toutMat),dims=1,na.rm=T)
# index of the hottest 10% of days
t90Idx <- which(tMean > quantile(tMean,0.9,na.rm=T))
# index (hour) of the max demand on the hottest days
t90kwhr = apply(data[t90Idx,],1,FUN=which.max)
# index (hour) of the min demand on the hottest days
t90kwhr_min = apply(data[t90Idx,],1,FUN=which.min)
# index (hour) of the max temperature on the hottest days
t90thr = apply(meterData$toutMat[t90Idx,],1,FUN=which.max)
# index (hour) of the min temperature on the hottest days
t90thr_min = apply(meterData$toutMat[t90Idx,],1,FUN=which.min)
# index of the coldest 10% of days
t10Idx <- which(tMean < quantile(tMean,0.1,na.rm=T))
# index (hour) of the max demand on the coldest days
t10kwhr = apply(data[t10Idx,],1,FUN=which.max)
# index (hour) of the max demand on the coldest days
t10kwhr_min = apply(data[t10Idx,],1,FUN=which.min)
# index (hour) of the max temperature on the coldest days
t10thr = apply(meterData$toutMat[t10Idx,],1,FUN=which.max)
# index (hour) of the min temperature on the coldest days
t10thr_min = apply(meterData$toutMat[t10Idx,],1,FUN=which.min)
# index of the top hours of consumption
kw90Idx = which(meterData$kw > quantile(meterData$kw,0.9,na.rm=T))
kw90hr = meterData$dates[kw90Idx]$hour
maxHOD <- as.numeric(which.max(hMean))
kw90 <- Mode(kw90hr)
t90kw <- Mode(t90kwhr)
t90kw_min <- Mode(t90kwhr_min)
t90t <- Mode(t90thr)
t90t_min <- Mode(t90thr_min)
t10kw <- Mode(t10kwhr)
t10kw_min <- Mode(t10kwhr_min)
t10t <- Mode(t10thr)
t10t_min <- Mode(t10thr_min)
dRange <- dMax - dMin
dHalfway <- dMin + (dMax - dMin) / 2
dHighD <- rowSums(data > dHalfway,dims=1)
dMn2mx <- dMin / dMax
dN2d <- rowMeans(data[,2:5],na.rm=T) / rowMeans(data[,16:19],na.rm=T) # 1-5am comp to 3-7pm
nv2dv <- var(as.vector(data[,2:5]),na.rm=T) / var(as.vector(data[,16:19]),na.rm=T) # var for night 1-5am comp to evening 3-7pm
# we need dates for this one...
#wkend = something to do with the DOW in the dates
#wkdays = ! wkend
#wknd2wk = mean(wkend) / mean(wkdays)
soft <- quantile(meterData$kw,c(0.03,0.97),na.rm=T)
max.97. <- soft[2] # this funky naming is inhereted from the name assigned when directly merging 'quantile' results into the list
min.3. <- soft[1] # this funky naming is inhereted from the name assigned when directly merging 'quantile' results into the list
nObs <- prod(dim(data))
kw.mean.annual = NA
enddt = as.POSIXlt(meterData$dates[length(meterData$dates)])
begindt = enddt
begindt$year = begindt$year - 1
lastYear = meterData$dates >= begindt
if(sum(lastYear,na.rm=T) > 365*24*0.9) { # observations 90% intact
#print('enough annual kw')
kw.mean.annual = mean(meterData$kw[lastYear],na.rm=T)
}
kw.mean = mean(meterData$kw,na.rm=T)
kw.mean.summer = mean(meterData$kw[ summerSubset],na.rm=T)
kw.mean.winter = mean(meterData$kw[!summerSubset],na.rm=T)
nObs.gas <- if(length(meterData$therms) > 0 ) { sum(!is.na(meterData$therms)) } else { 0 }
therm.mean.annual = NA
therm.mean = NA
therm.mean.summer = NA
therm.mean.winter = NA
therm.min = NA
date.first.gas = NA
date.last.gas = NA
if(length(meterData$therms > 0)) {
therm.mean = mean(meterData$therms,na.rm=T)
gasSummer = as.POSIXlt(meterData$gasDays)$mon %in% summerMon
therm.mean.summer = mean(meterData$therms[ gasSummer],na.rm=T)
therm.mean.winter = mean(meterData$therms[!gasSummer],na.rm=T)
thermdf = data.frame(therm=meterData$therms,day=meterData$gasDays,month=as.POSIXlt(meterData$gasDays)$mon)
thermMonth = reshape2::dcast(thermdf,month ~ .,mean,value.var='therm')
names(thermMonth) <- c('month','therm')
therm.min = mean(sort(thermMonth$therm,decreasing=F)[1:3]) # mean of three lowest months
#therm.min = mean(thermMonth$therm[thermMonth$month %in% 5:7]) # this is unreliable!!
date.first.gas = as.POSIXct(meterData$gasDays[1])
date.last.gas = as.POSIXct(meterData$gasDays[length(meterData$gasDays)])
if(nObs.gas >= 365) {
enddt = as.POSIXlt(meterData$gasDays[length(meterData$gasDays)])
begindt = enddt
begindt$year = begindt$year - 1
lastYear = meterData$gasDays >= begindt
if(sum(lastYear,na.rm=T) > 365*0.9) { # observations 90% intact
#print('enough annual therms')
therm.mean.annual = mean(meterData$therms[lastYear],na.rm=T)
}
}
}
maxIdx = which.max(meterData$kw)
minIdx = which.min(meterData$kw)
max.hr.kw = meterData$kw[maxIdx]
min.hr.kw = meterData$kw[minIdx]
max.hr.tout = meterData$tout[maxIdx]
min.hr.tout = meterData$tout[minIdx]
#print(meterData$dates[maxIdx])
max.hr.date = as.POSIXct(meterData$dates[maxIdx])
min.hr.date = as.POSIXct(meterData$dates[minIdx])
maxIdx = which.max(dMean)
minIdx = which.min(dMean)
max.day.kw = as.numeric(dMean[maxIdx] * 24)
min.day.kw = as.numeric(dMean[minIdx] * 24)
max.day.tout = tMean[maxIdx]
min.day.tout = tMean[minIdx]
max.day.date = meterData$days[maxIdx]
min.day.date = meterData$days[minIdx]
date.first = as.POSIXct(meterData$days[1])
date.last = as.POSIXct(meterData$days[length(meterData$days)])
# calculate percentiles of touts during max/min days
#print(max.day.tout)
#print(min.day.tout)
max.day.pct = mean(tMean <= max.day.tout)
min.day.pct = mean(tMean <= min.day.tout)
kw.tout.cor = cor(meterData$kw, meterData$tout, use='complete.obs') # correlation between tout and kw
kw.pout.cor = NA
#if(any(! is.na(meterData$w('pout')))) {
# kw.pout.cor = cor(meterData$kw,meterData$w('pout'), use='complete.obs') # correlation between pressure and kw
#}
kw.var = var(meterData$kw/kw.mean, use='complete.obs') # normed by mean of kW
kw.var.summer = var(meterData$kw[summerSubset]/kw.mean, na.rm=T) # normed by mean of kW
kw.var.winter = var(meterData$kw[!summerSubset]/kw.mean, na.rm=T) # normed by mean of kW
daily.kw.var = var(rowMeans(meterData$kwMat)/kw.mean, use='complete.obs') # normed by mean of kW
daily.kw.min.var = var(dMin/kw.mean, use='complete.obs') # normed by mean of kW
daily.kw.max.var = var(dMax/kw.mean, use='complete.obs') # normed by mean of kW
lags = 0:24
lag.cor = apply(as.matrix(lags), 1,function(x) cor(meterData$kw,lag(meterData$tout,x),use='complete.obs'))
lag.ma = apply(as.matrix(lags[-1]),1,function(x) cor(meterData$kw,ma(meterData$tout,x), use='complete.obs')) # no width 0
names(lag.cor) <- c(paste('lag',lags, sep=''))
lag.cor = as.list(lag.cor)
max.MA = which.max(lag.ma)
names(lag.ma) <- c(paste('ma', lags[-1],sep=''))
lag.ma = as.list(lag.ma)
basics = list(id=id,
nObs=nObs,
nObs.gas=nObs.gas,
kw.mean=kw.mean,
kw.mean.summer=kw.mean.summer,
kw.mean.winter=kw.mean.winter,
kw.mean.annual=kw.mean.annual,
therm.mean.annual=therm.mean.annual,
kw.total=kw.mean * 365 * 24,
therm.mean=therm.mean,
therm.mean.summer=therm.mean.summer,
therm.mean.winter=therm.mean.winter,
therm.total=therm.mean * 365 * 24,
therm.min=therm.min,
max.97.=max.97., # "soft" maximum at the 97th percentile
min.3.=min.3., # "soft" minimum at the 3rd percentile
kw.var=kw.var,
kw.var.summer=kw.var.summer,
kw.var.winter=kw.var.winter,
max.hr.kw=max.hr.kw,
min.hr.kw=min.hr.kw,
max.hr.tout=max.hr.tout,
min.hr.tout=min.hr.tout,
max.hr.date=max.hr.date,
min.hr.date=min.hr.date,
max.day.kw=max.day.kw,
min.day.kw=min.day.kw,
max.day.tout=max.day.tout,
min.day.tout=min.day.tout,
max.day.date=max.day.date,
min.day.date=min.day.date,
max.day.pct=max.day.pct,
min.day.pct=min.day.pct,
t90kw=t90kw,
t90kw_min=t90kw_min,
t90t=t90t,
t90t_min=t90t_min,
t10kw=t10kw,
t10kw_min=t10kw_min,
t10t=t10t,
t10t_min=t10t_min,
maxHOD=maxHOD,
kw90=kw90,
daily.kw.var=daily.kw.var,
daily.kw.min.var=daily.kw.min.var,
daily.kw.max.var=daily.kw.max.var,
kw.pout.cor=kw.pout.cor,
kw.tout.cor=kw.tout.cor,
max.MA=max.MA,
nv2dv=nv2dv,
date.first=date.first,
date.last=date.last,
date.first.gas=date.first.gas,
date.last.gas=date.last.gas
)
#print(class(basics))
dailyFeatures <- cbind(dMax,dMin,dMean,dRange,dHighD,dMn2mx,dN2d)
dailyFeatures[dailyFeatures == Inf] = NA # these will be caught by the na.rm in the mean fn
colnames(dailyFeatures) <- c('max','min','mean','range','dur','mn2mx','n2d')
monthFeatures = t(sapply(split(data.frame(dailyFeatures),factor(format(meterData$days,'%b'),levels=month.abb)),colMeans,na.rm=T))
# return the means across all days
augF = monthFeatures[c('Aug'),]
names(augF) <- paste('Aug_',names(augF),sep='')
janF = monthFeatures[c('Jan'),]
names(janF) <- paste('Jan_',names(janF),sep='')
meanDailyFeatures = as.list(apply(dailyFeatures,2,FUN=mean,na.rm=T))
featureList <- c(basics, meanDailyFeatures, monthTotals, hMean, lag.cor, lag.ma, augF, janF)
#print(featureList)
return(featureList)
}
ConvergenceDA/visdom documentation built on May 6, 2019, 12:51 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Defines functions basicFeatures
Documented in basicFeatures
# Copyright 2016 The Board of Trustees of the Leland Stanford Junior University.
# Direct inquiries to Sam Borgeson (sborgeson@stanford.edu)
# or professor Ram Rajagopal (ramr@stanford.edu)
#' @title
#' Implementation of several basic statistical features
#'
#' @description
#' Function that implements a set of common statistical features that can be run on an instance of \code{\link{MeterDataClass}}.
#'
#' @param meterData An instance of \code{\link{MeterDataClass}} that provides the data used for all the implemented feature calculations.
#'
#' @details
#' \code{basicFeatures} is called by passing in an instance of a MeterDataClass, which provides the meter data in both linear series
#' and day-per-row matrix formats. These data structures and some associated indices and intermediate data structures are
#' re-used throughout this method, which tries to take advantage of the performance benefits of such re-use.
#' This function is often called in the context of the \code{iterator.*} suite of functions, which can loop through large sets of
#' meters, calling feature functions on each.
#'
#' @examples
#' \dontrun{
#' DATA_SOURCE = TestData(n=10)
#' meterData = DATA_SOURCE$getMeterDataClass(DATA_SOURCE$getIds()[1])
#' features = basicFeatures(meterData)
#' names(features)
#' class(features)
#' }
#' @export
basicFeatures = function(meterData, ...){ # r is an instance of MeterDataClass
data <- as.matrix(meterData$kwMat)
id <- meterData$id
monthTotals = sapply(split(meterData$kw,factor(format(meterData$dates,'%b'),levels=month.abb)),mean,na.rm=T) * 30 * 24
names(monthTotals) <- paste('kw.total.',names(monthTotals),sep='')
monthTotals = as.list(monthTotals) # needs to be a list ofr concatenation to work
summerMon = 4:8 # May through Sept - zero based
summerSubset = as.POSIXlt(meterData$dates)$mon %in% summerMon
wkdySubset = as.POSIXlt(meterData$dates)$day %in% 1:5
# todo: apply/quantile is slow. Is it worth softening min,max this way?
qtle <- apply(data,1,FUN=quantile,c(0.01,0.99),na.rm=T)
dMin <- qtle[1,]
dMax <- qtle[2,]
hMean <- colMeans(data,na.rm=T)
names(hMean) <- paste('HOD.mean.',1:24,sep='')
dMean <- rowMeans(data,dims=1,na.rm=T) # rowMeans is faster than apply with FUN=mean ?rowMeans for details
tMean <- rowMeans(as.matrix(meterData$toutMat),dims=1,na.rm=T)
# index of the hottest 10% of days
t90Idx <- which(tMean > quantile(tMean,0.9,na.rm=T))
# index (hour) of the max demand on the hottest days
t90kwhr = apply(data[t90Idx,],1,FUN=which.max)
# index (hour) of the min demand on the hottest days
t90kwhr_min = apply(data[t90Idx,],1,FUN=which.min)
# index (hour) of the max temperature on the hottest days
t90thr = apply(meterData$toutMat[t90Idx,],1,FUN=which.max)
# index (hour) of the min temperature on the hottest days
t90thr_min = apply(meterData$toutMat[t90Idx,],1,FUN=which.min)
# index of the coldest 10% of days
t10Idx <- which(tMean < quantile(tMean,0.1,na.rm=T))
# index (hour) of the max demand on the coldest days
t10kwhr = apply(data[t10Idx,],1,FUN=which.max)
# index (hour) of the max demand on the coldest days
t10kwhr_min = apply(data[t10Idx,],1,FUN=which.min)
# index (hour) of the max temperature on the coldest days
t10thr = apply(meterData$toutMat[t10Idx,],1,FUN=which.max)
# index (hour) of the min temperature on the coldest days
t10thr_min = apply(meterData$toutMat[t10Idx,],1,FUN=which.min)
# index of the top hours of consumption
kw90Idx = which(meterData$kw > quantile(meterData$kw,0.9,na.rm=T))
kw90hr = meterData$dates[kw90Idx]$hour
maxHOD <- as.numeric(which.max(hMean))
kw90 <- Mode(kw90hr)
t90kw <- Mode(t90kwhr)
t90kw_min <- Mode(t90kwhr_min)
t90t <- Mode(t90thr)
t90t_min <- Mode(t90thr_min)
t10kw <- Mode(t10kwhr)
t10kw_min <- Mode(t10kwhr_min)
t10t <- Mode(t10thr)
t10t_min <- Mode(t10thr_min)
dRange <- dMax - dMin
dHalfway <- dMin + (dMax - dMin) / 2
dHighD <- rowSums(data > dHalfway,dims=1)
dMn2mx <- dMin / dMax
dN2d <- rowMeans(data[,2:5],na.rm=T) / rowMeans(data[,16:19],na.rm=T) # 1-5am comp to 3-7pm
nv2dv <- var(as.vector(data[,2:5]),na.rm=T) / var(as.vector(data[,16:19]),na.rm=T) # var for night 1-5am comp to evening 3-7pm
# we need dates for this one...
#wkend = something to do with the DOW in the dates
#wkdays = ! wkend
#wknd2wk = mean(wkend) / mean(wkdays)
soft <- quantile(meterData$kw,c(0.03,0.97),na.rm=T)
max.97. <- soft[2] # this funky naming is inhereted from the name assigned when directly merging 'quantile' results into the list
min.3. <- soft[1] # this funky naming is inhereted from the name assigned when directly merging 'quantile' results into the list
nObs <- prod(dim(data))
kw.mean.annual = NA
enddt = as.POSIXlt(meterData$dates[length(meterData$dates)])
begindt = enddt
begindt$year = begindt$year - 1
lastYear = meterData$dates >= begindt
if(sum(lastYear,na.rm=T) > 365*24*0.9) { # observations 90% intact
#print('enough annual kw')
kw.mean.annual = mean(meterData$kw[lastYear],na.rm=T)
}
kw.mean = mean(meterData$kw,na.rm=T)
kw.mean.summer = mean(meterData$kw[ summerSubset],na.rm=T)
kw.mean.winter = mean(meterData$kw[!summerSubset],na.rm=T)
nObs.gas <- if(length(meterData$therms) > 0 ) { sum(!is.na(meterData$therms)) } else { 0 }
therm.mean.annual = NA
therm.mean = NA
therm.mean.summer = NA
therm.mean.winter = NA
therm.min = NA
date.first.gas = NA
date.last.gas = NA
if(length(meterData$therms > 0)) {
therm.mean = mean(meterData$therms,na.rm=T)
gasSummer = as.POSIXlt(meterData$gasDays)$mon %in% summerMon
therm.mean.summer = mean(meterData$therms[ gasSummer],na.rm=T)
therm.mean.winter = mean(meterData$therms[!gasSummer],na.rm=T)
thermdf = data.frame(therm=meterData$therms,day=meterData$gasDays,month=as.POSIXlt(meterData$gasDays)$mon)
thermMonth = reshape2::dcast(thermdf,month ~ .,mean,value.var='therm')
names(thermMonth) <- c('month','therm')
therm.min = mean(sort(thermMonth$therm,decreasing=F)[1:3]) # mean of three lowest months
#therm.min = mean(thermMonth$therm[thermMonth$month %in% 5:7]) # this is unreliable!!
date.first.gas = as.POSIXct(meterData$gasDays[1])
date.last.gas = as.POSIXct(meterData$gasDays[length(meterData$gasDays)])
if(nObs.gas >= 365) {
enddt = as.POSIXlt(meterData$gasDays[length(meterData$gasDays)])
begindt = enddt
begindt$year = begindt$year - 1
lastYear = meterData$gasDays >= begindt
if(sum(lastYear,na.rm=T) > 365*0.9) { # observations 90% intact
#print('enough annual therms')
therm.mean.annual = mean(meterData$therms[lastYear],na.rm=T)
}
}
}
maxIdx = which.max(meterData$kw)
minIdx = which.min(meterData$kw)
max.hr.kw = meterData$kw[maxIdx]
min.hr.kw = meterData$kw[minIdx]
max.hr.tout = meterData$tout[maxIdx]
min.hr.tout = meterData$tout[minIdx]
#print(meterData$dates[maxIdx])
max.hr.date = as.POSIXct(meterData$dates[maxIdx])
min.hr.date = as.POSIXct(meterData$dates[minIdx])
maxIdx = which.max(dMean)
minIdx = which.min(dMean)
max.day.kw = as.numeric(dMean[maxIdx] * 24)
min.day.kw = as.numeric(dMean[minIdx] * 24)
max.day.tout = tMean[maxIdx]
min.day.tout = tMean[minIdx]
max.day.date = meterData$days[maxIdx]
min.day.date = meterData$days[minIdx]
date.first = as.POSIXct(meterData$days[1])
date.last = as.POSIXct(meterData$days[length(meterData$days)])
# calculate percentiles of touts during max/min days
#print(max.day.tout)
#print(min.day.tout)
max.day.pct = mean(tMean <= max.day.tout)
min.day.pct = mean(tMean <= min.day.tout)
kw.tout.cor = cor(meterData$kw, meterData$tout, use='complete.obs') # correlation between tout and kw
kw.pout.cor = NA
#if(any(! is.na(meterData$w('pout')))) {
# kw.pout.cor = cor(meterData$kw,meterData$w('pout'), use='complete.obs') # correlation between pressure and kw
#}
kw.var = var(meterData$kw/kw.mean, use='complete.obs') # normed by mean of kW
kw.var.summer = var(meterData$kw[summerSubset]/kw.mean, na.rm=T) # normed by mean of kW
kw.var.winter = var(meterData$kw[!summerSubset]/kw.mean, na.rm=T) # normed by mean of kW
daily.kw.var = var(rowMeans(meterData$kwMat)/kw.mean, use='complete.obs') # normed by mean of kW
daily.kw.min.var = var(dMin/kw.mean, use='complete.obs') # normed by mean of kW
daily.kw.max.var = var(dMax/kw.mean, use='complete.obs') # normed by mean of kW
lags = 0:24
lag.cor = apply(as.matrix(lags), 1,function(x) cor(meterData$kw,lag(meterData$tout,x),use='complete.obs'))
lag.ma = apply(as.matrix(lags[-1]),1,function(x) cor(meterData$kw,ma(meterData$tout,x), use='complete.obs')) # no width 0
names(lag.cor) <- c(paste('lag',lags, sep=''))
lag.cor = as.list(lag.cor)
max.MA = which.max(lag.ma)
names(lag.ma) <- c(paste('ma', lags[-1],sep=''))
lag.ma = as.list(lag.ma)
basics = list(id=id,
nObs=nObs,
nObs.gas=nObs.gas,
kw.mean=kw.mean,
kw.mean.summer=kw.mean.summer,
kw.mean.winter=kw.mean.winter,
kw.mean.annual=kw.mean.annual,
therm.mean.annual=therm.mean.annual,
kw.total=kw.mean * 365 * 24,
therm.mean=therm.mean,
therm.mean.summer=therm.mean.summer,
therm.mean.winter=therm.mean.winter,
therm.total=therm.mean * 365 * 24,
therm.min=therm.min,
max.97.=max.97., # "soft" maximum at the 97th percentile
min.3.=min.3., # "soft" minimum at the 3rd percentile
kw.var=kw.var,
kw.var.summer=kw.var.summer,
kw.var.winter=kw.var.winter,
max.hr.kw=max.hr.kw,
min.hr.kw=min.hr.kw,
max.hr.tout=max.hr.tout,
min.hr.tout=min.hr.tout,
max.hr.date=max.hr.date,
min.hr.date=min.hr.date,
max.day.kw=max.day.kw,
min.day.kw=min.day.kw,
max.day.tout=max.day.tout,
min.day.tout=min.day.tout,
max.day.date=max.day.date,
min.day.date=min.day.date,
max.day.pct=max.day.pct,
min.day.pct=min.day.pct,
t90kw=t90kw,
t90kw_min=t90kw_min,
t90t=t90t,
t90t_min=t90t_min,
t10kw=t10kw,
t10kw_min=t10kw_min,
t10t=t10t,
t10t_min=t10t_min,
maxHOD=maxHOD,
kw90=kw90,
daily.kw.var=daily.kw.var,
daily.kw.min.var=daily.kw.min.var,
daily.kw.max.var=daily.kw.max.var,
kw.pout.cor=kw.pout.cor,
kw.tout.cor=kw.tout.cor,
max.MA=max.MA,
nv2dv=nv2dv,
date.first=date.first,
date.last=date.last,
date.first.gas=date.first.gas,
date.last.gas=date.last.gas
)
#print(class(basics))
dailyFeatures <- cbind(dMax,dMin,dMean,dRange,dHighD,dMn2mx,dN2d)
dailyFeatures[dailyFeatures == Inf] = NA # these will be caught by the na.rm in the mean fn
colnames(dailyFeatures) <- c('max','min','mean','range','dur','mn2mx','n2d')
monthFeatures = t(sapply(split(data.frame(dailyFeatures),factor(format(meterData$days,'%b'),levels=month.abb)),colMeans,na.rm=T))
# return the means across all days
augF = monthFeatures[c('Aug'),]
names(augF) <- paste('Aug_',names(augF),sep='')
janF = monthFeatures[c('Jan'),]
names(janF) <- paste('Jan_',names(janF),sep='')
meanDailyFeatures = as.list(apply(dailyFeatures,2,FUN=mean,na.rm=T))
featureList <- c(basics, meanDailyFeatures, monthTotals, hMean, lag.cor, lag.ma, augF, janF)
#print(featureList)
return(featureList)
}
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.