R简单的使用使用霍尔特指数平滑法预测数据

1、连接oracle数据库，提取所需数据

library(DBI)

library(ROracle)

library(forecast)

drv=dbDriver('Oracle')

conn=dbConnect(drv,'trff_app','trff_admin','192.111.11.00:1521/jgyw')

rs=dbSendQuery(conn,'select  WFS,GLBM,WFYF from vio_violation_glbmyc where wfyf>to_number(extract(year from sysdate)-2||12)')

data=fetch(rs)

2、第一次用R生成数据表值下一年月的笨方法

#next yf begin

nextyf_n=dbSendQuery(conn,'select extract(year from sysdate) nexty_n from dual ')

nextyf_y=dbSendQuery(conn,'select extract(month from sysdate) nexty_y from dual ')

nextyf_n=fetch(nextyf_n)

nextyf_y=fetch(nextyf_y)

nextyf_n=nextyf_n$NEXTY_N

nextyf_y=nextyf_y$NEXTY_Y

if(nextyf_y<11)

  nextyf_y=paste(0,nextyf_y,sep = "")

if(nextyf_y==12)

{nextyf_y=paste(0,1,sep = "")

nextyf_n=nextyf_n+1

}

nextyf=paste(nextyf_n,nextyf_y,sep ='' )

#end

简单的方法

dbSendQuery直接用sql语句生成，但是注意select语句要用双引号

3、数据准备，next

rsjkdbh=dbSendQuery(conn,'select distinct GLBM from vio_violation_glbmyc where wfyf>to_number(extract(year from sysdate)-2||12)')

datajkdbh=fetch(rsjkdbh)

#datajkdbh=as.matrix(datajkdbh)

#datajkdbh=datajkdbh[order(datajkdbh[,1]),]

jkdlength=nrow(datajkdbh)

rswfyf=dbSendQuery(conn,'select distinct wfyf from vio_violation_glbmyc where wfyf>to_number(extract(year from sysdate)-2||12)')

datawfyf=fetch(rswfyf)

datawfyf=as.matrix(datawfyf)

datawfyf=as.numeric(datawfyf[order(datawfyf[,1]),])

datawfyf=t(t(datawfyf))

yfrqlength=nrow(datawfyf)

datarow=0

datajkdtemp=0

datanextv=0

i=0

4、使用HoltWinters法进行数据预测，如果历史数据较少则走else

for (i in 1:jkdlength)

{

  #i=28

  datarow=data[data[,2]==datajkdbh[i,],]

  if(nrow(datarow)>2)

  {

    datajkdtemp=datajkdbh[i,]

    datarow=datarow[sort(datarow$WFYF,index.return=TRUE)$ix,]

    seqvalue=as.numeric(as.matrix(datarow$WFS))

    #skirtsseries<-log(seqvalue)+1

    skirtsseries<-ts(seqvalue,start=c(1))

    

    #skirtsseriesforecasts<-HoltWinters(skirtsseries,gamma=FALSE)

    skirtsseriesforecasts<-HoltWinters(skirtsseries,beta=FALSE,gamma=FALSE)

    skirtsseriesforecasts2<-forecast.HoltWinters(skirtsseriesforecasts,h=1)

    nextv=skirtsseriesforecasts2$mean[1]

    #nextv=round(exp(nextv-1))

    nextv=round(nextv)

    if(nextv<0)

      nextv=round(mean(seqvalue))

    retur=c(nextv,datajkdtemp,nextyf)

    retur=t(retur)

    retur=as.data.frame(retur)

    colnames(retur)=c('WFS','GLBM','WFYF')

    

    dbWriteTable(conn,'VIO_VIOLATION_GLBMYC',retur, row.names = F, append = TRUE)

  }

  else

  {

    datajkdtemp=datajkdbh[i,]

    datarow=datarow[sort(datarow$WFYF,index.return=TRUE)$ix,]

    seqvalue=as.numeric(as.matrix(datarow$WFS))

    

    nextv=round(mean(seqvalue))

    retur=c(nextv,datajkdtemp,nextyf)

    retur=t(retur)

    retur=as.data.frame(retur)

    colnames(retur)=c('WFS','GLBM','WFYF')

    

    dbWriteTable(conn,'VIO_VIOLATION_GLBMYC',retur, row.names = F, append = TRUE)

  }

  

  #dbWriteTable(conn,'zz_vio_veh_viorec_init',retur,overwrite=TRUE)

  

}

5、over（不喜勿碰）