业务临时需要&＃xff0c;需要确定多因素影响下的综合权重值&＃xff0c;现使用层次分析法和拉格朗日多项式插值算法做简易值计算。
1、建立层次分析结构模型&＃xff0c;分析影响综合指标的各个因素&＃xff0c;分层级&＃xff0c;上层受下层影响&＃xff0c;而同层各因素之间基本上相对独立(本实例只有两层)
2、连接oracle数据库&＃xff0c;读取各个因素实际指标值
library(DBI)
library(ROracle)
drv&＃61;dbDriver(&＃39;Oracle&＃39;)
conn&＃61;dbConnect(drv,&＃39;AQTS_ZHZX&＃39;,&＃39;AQTS_ZHZX&＃39;,&＃39;192.160.0.0:1521/iims&＃39;)
rs&＃61;dbSendQuery(conn,"with weather as
(
select * from (
select * from weather_fact_aqts w
where w.cjrq&＃61;(select max(cjrq) from weather_fact_aqts)
order by w.cjsd desc) where rownum&＃61;1
)
select v.yfbm,v.dlbm,v.ldbm,weather.cjrq,weather.cjsd,v.wfshb,a.sgshb,
weather.sbbh,weather.wd,weather.sd,weather.fl,weather.njd
,weather.fog_rank,weather.wind_rank from VIO_FACT_AQTS v
join ACD_FACT_AQTS a on v.yfbm&＃61;a.sgyfbm and v.dlbm&＃61;a.lhbm and v.ldbm&＃61;a.glsbm
join weather on 1&＃61;1")
data&＃61;fetch(rs)
data&＃61;as.matrix(data)
3、影响综合因素的指标分两级&＃xff1a;
#FIRST,A1:weather,A2:Flux,A3:acd,A4:vio
#SECOND1,B1:fog,B2:rain,B3:wind,B4:snow
#SECOND2,B2:FLUX1&＃xff0c;FLUX2
4、构造第一层成对比较矩阵(采用成对比较法和1~9尺度)
#FIRST,A1:weather,A2:Flux,A3:acd,A4:vio
A&＃61;matrix(0,4,4)
diag(A)&＃61;1
A[1,2]&＃61;1/2;A[2,1]&＃61;2
A[1,3]&＃61;1/4;A[3,1]&＃61;4
A[1,4]&＃61;1/4;A[4,1]&＃61;4
A[2,4]&＃61;1/2;A[4,2]&＃61;2
A[2,3]&＃61;1/2;A[3,2]&＃61;2
A[3,4]&＃61;A[4,3]&＃61;1
5、构造第二层成对比较矩阵(构造针对上一层成对比较矩阵)
#SECOND,B1:fog,B2:rain,B3:wind,B4:snow
B&＃61;matrix(0,4,4)
diag(B)&＃61;1
B[1,2]&＃61;2;B[2,1]&＃61;1/2
B[1,3]&＃61;6;B[3,1]&＃61;1/6
B[1,4]&＃61;2;B[4,1]&＃61;1/2
B[2,3]&＃61;3;B[3,2]&＃61;1/3
B[2,4]&＃61;1;B[4,2]&＃61;1
B[3,4]&＃61;1/3;B[4,3]&＃61;3
#SECOND2,B2:FLUX1&＃xff0c;FLUX2
C&＃61;matrix(1,2,2)
C[1,2]&＃61;3
C[2,1]&＃61;1/3
6、计算权向量并一致性校验
求权向量以及特征向量&＃xff0c;进行一致性检验(因时间问题&＃xff0c;这块没有加校验&＃xff0c;后续修改)
7、求各层级因素权重&＃xff0c;这里使用规范列平均法
Ac&＃61;colSums(A)
for (i in 1:nrow(A)) {
A[,i]&＃61;A[,i]/Ac[i]
}
weigA&＃61;rowMeans(A)
#B
Bc&＃61;colSums(B)
for (i in 1:nrow(B)) {
B[,i]&＃61;B[,i]/Bc[i]
}
weigB&＃61;rowMeans(B)
#C 1:flux,2:vehicle
Cc&＃61;colSums(C)
for (i in 1:nrow(C)) {
C[,i]&＃61;C[,i]/Cc[i]
}
weigC&＃61;rowMeans(C)
########row 1:fog,2:rain,3:wind,4:snow,5:flux,6:vehicle,7:acd,8:vio
########new row 1:fog,2:rain,3:wind,4:snow,5:even,6:acd,7:vio
weit&＃61;matrix(0,8,4)
weit[1:4,1]&＃61;weigB
weit[5:6,2]&＃61;weigC
weit[7,3]&＃61;1
weit[8,4]&＃61;1
######
w&＃61;as.matrix(weigA)
p&＃61;weit
we&＃61;p%*%w
wenew&＃61;c(we[1:4],sum(we[5:6]),we[7],we[8])
8、构造各个因素的函数
#get the function of accident
acdx
acdy
acdf
acdv
x&＃61;as.numeric(x)
ranka&＃61;f(x)
ranka[ranka>5]&＃61;5
ranka[ranka<0]&＃61;0
return (ranka)
}
#v&＃61;c(-0.1,0.2)
#vv&＃61;acdv(v,acdf)
#get the function of violation
viox
vioy
viof
#get the function of flux flux
fluxfx
fluxfy
fluxff
#get the function of flux vehicel
fluxvx
fluxvy
fluxvf
#get the function of fog rain snow
fogx
fogy
fogf
#get the function of wind
windx
windy
windf
9、对应各因素指标值与权重相乘求取综合指标级别&＃xff0c;并写入数据库
rankc&＃61;cbind(data[,c(2:5)],as.matrix(acdv(data[,12],fogf))*wenew[1],as.matrix(acdv(data[,12],fogf))*wenew[2],
as.matrix(acdv(data[,11],fogf))*wenew[3],as.matrix(acdv(data[,12],fogf))*wenew[4],as.matrix(rep(3,nrow(data)))*wenew[5],
as.matrix(acdv(data[,7],acdf))*wenew[6],as.matrix(acdv(data[,6],viof))*wenew[7])
#comprehensive rank
alv&＃61;as.numeric(rankc[,5:11])
dim(alv)&＃61;c(nrow(rankc),7)
RANKOVER
colnames(RANKOVER)
RANKOVER&＃61;as.data.frame(RANKOVER)
#dbRemoveTable(conn, &＃39;ZHZX_FACT_AQTS&＃39;)
dbWriteTable(conn,&＃39;ZHZX_FACT_AQTS&＃39;,RANKOVER, row.names &＃61; F, append &＃61; TRUE)
附&＃xff1a;拉格朗日多项式插值算法
#LagrangePolynomial
LagrangePolynomial
len &＃61; length(x)
if(len !&＃61; length(y))
stop("length not equal!")
if(len
stop("dim size must more than 1")
#pretreat data abd alloc memery
xx
m
#combin express
for(i in 1:len) {
td
tm
for(j in 1:len) {
if(i !&＃61; j) {
td
tm
}
}
tm
m[i]
}
#combin the exrpession
m
r
#combin the function
fbody
f
#fill the function&＃39;s body
body(f)
return(f)
}
(记录日常点滴)对算法以及R使用还不是很熟悉&＃xff0c;后续完善&＃xff01;
如有错误&＃xff0c;希望指点>>>OK
京公网安备 11010802041100号