====== Clamix / Cars ======
[[notes:food|Food]]
[[notes:clamix|Specificity]]
===== Transforming car data into SO =====
The raw data were obtained from **Cars Catalog 1997** based on //Katalog Avtomobilov '97 / Posebna priloga Dela in Slovenskih novic April '97// (by Janko Blagojevič). Transformation into symbolic objects (SOs) by Vladimir Batagelj, 29. July 2010.
The scheme of analysis is described in ''analysis.R'' (see also ''trace.txt''). The cars data set is already transformed in symbolic objects (''cars.so'' and ''cars.meta''). To recreate or change transformation see ''makeCars.R''.
The relevant files are collected in {{:pub:zip:cars97.zip}} .
cars.csv - cars data set
cars.nam - short and long names of cars
encCars.R - encoding rules for the cars data set
makeCars.R - creation of cars.so and cars.meta
analysis.R - example steps in analysis
trace.txt - trace of analysis
cars.Rdata - encoded cars data
cars.so - cars data set transformed in symbolic objects
cars.meta - metadata for cars symbolic objects - needed for
post-analyses
cars.rez - results of leaders method
dendro.pdf - dendrogram from hierarchical clustering
See also {{:paper:clamix:clamixrs.pdf|34th GfKl}} / 21-23. July 2010 (slides 13-16) and (Nataša's approach) {{:paper:clamix:sredin1187_v2.pdf|Wednesday seminar 1187}} / 27. October 2010 (slides 23-27).
==== Details ====
For transforming the data we have to prepare the encoding rules (file ''encCars.R'') for all variables:
# V. Batagelj, 29.7.2010
# -------------------------------------------------------------------
encPrice <- list(
"[950,2100]" = function(x) x<=2100,
"(2100,2800]" = function(x) x<=2800,
"(2800,3700]" = function(x) x<=3700,
"(3700,5000]" = function(x) x<=5000,
"(5000,6550]" = function(x) x<=6550,
"(6550,45000]" = function(x) x<=45000,
"NA" = function(x) TRUE )
levType <- c("LI","KL","EN","KA","KB","RO","TE","KU")
namesNumDoors <- c(2:5,"NA")
codeNumDoors <- function(x){
if(is.na(x)) return(5)
if((2<=x)&&(x<=5)) return(x-1)
return(0)
}
namesNumPassen <- c(2:8,"NA")
codeNumPassen <- function(x){
if(is.na(x)) return(8)
if((2<=x)&&(x<=8)) return(x-1)
return(0)
}
levMsite <- c("S","Z","SR")
levDrive <- c("S","Z","S4","E")
encLen <- list(
"[2600,4010]" = function(x) x<=4010,
"(4010,4245]" = function(x) x<=4245,
"(4245,4470]" = function(x) x<=4470,
"(4470,4555]" = function(x) x<=4555,
"(4555,4761]" = function(x) x<=4761,
"(4761,6000]" = function(x) x<=6000,
"NA" = function(x) TRUE )
encWid <- list(
"[1350,1680]" = function(x) x<=1680,
"(1680,1700]" = function(x) x<=1700,
"(1700,1735]" = function(x) x<=1735,
"(1735,1780]" = function(x) x<=1780,
"(1780,1812]" = function(x) x<=1812,
"(1812,2000]" = function(x) x<=2000,
"NA" = function(x) TRUE )
encHei <- list(
"[1045,1388]" = function(x) x<=1388,
"(1388,1410]" = function(x) x<=1410,
"(1410,1420]" = function(x) x<=1420,
"(1420,1438]" = function(x) x<=1438,
"(1438,1490]" = function(x) x<=1490,
"(1490,2500]" = function(x) x<=2500,
"NA" = function(x) TRUE )
encWba <- list(
"[1450,2445]" = function(x) x<=2445,
"(2445,2510]" = function(x) x<=2510,
"(2510,2590]" = function(x) x<=2590,
"(2590,2680]" = function(x) x<=2680,
"(2680,2730]" = function(x) x<=2730,
"(2730,3500]" = function(x) x<=3500,
"NA" = function(x) TRUE )
encLug <- list(
"[75,279]" = function(x) x<=279,
"(279,360]" = function(x) x<=360,
"(360,429]" = function(x) x<=429,
"(429,480]" = function(x) x<=480,
"(480,560]" = function(x) x<=560,
"(560,5000]" = function(x) x<=5000,
"NA" = function(x) TRUE )
levELu <- c("da","ne")
encFCa <- list(
"[30,48]" = function(x) x<=48,
"(48,53]" = function(x) x<=53,
"(53,60]" = function(x) x<=60,
"(60,66]" = function(x) x<=66,
"(66,76]" = function(x) x<=76,
"(76,150]" = function(x) x<=150,
"NA" = function(x) TRUE )
encWei <- list(
"[225,940]" = function(x) x<=940,
"(940,1080]" = function(x) x<=1080,
"(1080,1220]" = function(x) x<=1220,
"(1220,1350]" = function(x) x<=1350,
"(1350,1540]" = function(x) x<=1540,
"(1540,3000]" = function(x) x<=3000,
"NA" = function(x) TRUE )
encMlo <- list(
"[70,455]" = function(x) x<=455,
"(455,490]" = function(x) x<=490,
"(490,515]" = function(x) x<=515,
"(515,550]" = function(x) x<=550,
"(550,650]" = function(x) x<=650,
"(650,3000]" = function(x) x<=3000,
"NA" = function(x) TRUE )
encDsp <- list(
"[796,1490]" = function(x) x<=1490,
"(1490,1690]" = function(x) x<=1690,
"(1690,1886]" = function(x) x<=1886,
"(1886,2000]" = function(x) x<=2000,
"(2000,2500]" = function(x) x<=2500,
"(2500,6000]" = function(x) x<=6000,
"NA" = function(x) TRUE )
encMkw <- list(
"[26,55]" = function(x) x<=55,
"(55,66]" = function(x) x<=66,
"(66,85]" = function(x) x<=85,
"(85,99]" = function(x) x<=99,
"(99,125]" = function(x) x<=125,
"(125,400]" = function(x) x<=400,
"NA" = function(x) TRUE )
encMkm <- list(
"[25,75]" = function(x) x<=75,
"(75,99]" = function(x) x<=99,
"(99,118]" = function(x) x<=118,
"(118,144]" = function(x) x<=144,
"(144,200]" = function(x) x<=200,
"(200,500]" = function(x) x<=500,
"NA" = function(x) TRUE )
encRMp <- list(
"[3400,4500]" = function(x) x<=4500,
"(4500,5250]" = function(x) x<=5250,
"(5250,5500]" = function(x) x<=5500,
"(5500,5700]" = function(x) x<=5700,
"(5700,6000]" = function(x) x<=6000,
"(6000,9000]" = function(x) x<=9000,
"NA" = function(x) TRUE )
encMtq <- list(
"[55,124]" = function(x) x<=124,
"(124,145]" = function(x) x<=145,
"(145,172]" = function(x) x<=172,
"(172,202]" = function(x) x<=202,
"(202,270]" = function(x) x<=270,
"(270,600]" = function(x) x<=600,
"NA" = function(x) TRUE )
encRMt <- list(
"[1600,2250]" = function(x) x<=2250,
"(2250,2950]" = function(x) x<=2950,
"(2950,3500]" = function(x) x<=3500,
"(3500,4000]" = function(x) x<=4000,
"(4000,4500]" = function(x) x<=4500,
"(4500,7000]" = function(x) x<=7000,
"NA" = function(x) TRUE )
levTrans <- c("A3","A4","A5","R4","R5","R5*","R6","S5","T4","T5","RS")
levBreak <- c("K/B","K/K")
encFue <- list(
"[3.6,5]" = function(x) x<=5,
"(5,5.6]" = function(x) x<=5.6,
"(5.6,6.1]" = function(x) x<=6.1,
"(6.1,6.9]" = function(x) x<=6.9,
"(6.9,8]" = function(x) x<=8,
"(8,20]" = function(x) x<=20,
"NA" = function(x) TRUE )
encAcc <- list(
"[4.4,9.1]" = function(x) x<=9.1,
"(9.1,10.6]" = function(x) x<=10.6,
"(10.6,11.6]" = function(x) x<=11.6,
"(11.6,12.9]" = function(x) x<=12.9,
"(12.9,15.3]" = function(x) x<=15.3,
"(15.3,32]" = function(x) x<=32,
"NA" = function(x) TRUE )
encSpe <- list(
"[130,163]" = function(x) x<=163,
"(163,174]" = function(x) x<=174,
"(174,187]" = function(x) x<=187,
"(187,200]" = function(x) x<=200,
"(200,215]" = function(x) x<=215,
"(215,400]" = function(x) x<=400,
"NA" = function(x) TRUE )
The file ''Clamix3.R'' contains some functions supporting the encoding:
# creates an empty SO
emptySO <- function(nCats){
nVar <- length(nCats)
s <- vector("list", nVar)
for(i in 1:nVar) s[[i]] <- double(nCats[i]+1)
return(s)
}
# encode numerical vector using given encoding
encodeSO <- function(x,encoding,codeNA){
if(is.na(x)) return(codeNA)
for(i in 1:length(encoding)) if(encoding[[i]](x)) return(i)
}
The encoding is done by commands in the file ''makeCars.R'':
setwd("C:/Users/Batagelj/work/clamix/clamix.R")
source("C:\\Users\\Batagelj\\work\\clamix\\clamix.R\\cars\\encCars.R")
source("C:\\Users\\Batagelj\\work\\clamix\\clamix.R\\clamix3.R")
cf <- read.table("./cars/cars.csv",header=TRUE,dec=".",row.names=1)
numSO <- nrow(cf)
nVar <- ncol(cf); nVarP <- nVar+1
carsSO <- vector("list",nVar)
names(carsSO) <- colnames(cf)
carsCats <- vector("list",nVar)
names(carsCats) <- colnames(cf)
carsCats$price <- names(encPrice)
carsSO$price <- sapply(cf$price,function(x) encodeSO(x,encPrice,7))
carsCats$type <- c(levType,NA)
carsSO$type <- as.integer(factor(cf$type,levels=levType))
carsCats$NumDoors <- namesNumDoors
carsSO$NumDoors <- sapply(cf$NumDoors,codeNumDoors)
carsCats$NumPassen <- namesNumPassen
carsSO$NumPassen <- sapply(cf$NumPassen,codeNumPassen)
carsCats$motorsite <- c(levMsite,NA)
carsSO$motorsite <- as.integer(factor(cf$motorsite,levels=levMsite))
carsCats$drive <- c(levDrive,NA)
carsSO$drive <- as.integer(factor(cf$drive,levels=levDrive))
carsCats$length <- names(encLen)
carsSO$length <- sapply(cf$length,function(x) encodeSO(x,encLen,7))
carsCats$width <- names(encWid)
carsSO$width <- sapply(cf$width,function(x) encodeSO(x,encWid,7))
carsCats$height <- names(encHei)
carsSO$height <- sapply(cf$height,function(x) encodeSO(x,encHei,7))
carsCats$wheelbase <- names(encWba)
carsSO$wheelbase <- sapply(cf$wheelbase,function(x) encodeSO(x,encWba,7))
carsCats$luggage <- names(encLug)
carsSO$luggage <- sapply(cf$luggage,function(x) encodeSO(x,encLug,7))
carsCats$enlarLugg <- c(levELu,NA)
carsSO$enlarLugg <- as.integer(factor(cf$enlarLugg,levels=levELu))
carsCats$fuelCapac <- names(encFCa)
carsSO$fuelCapac <- sapply(cf$fuelCapac,function(x) encodeSO(x,encFCa,7))
carsCats$weight <- names(encWei)
carsSO$weight <- sapply(cf$weight,function(x) encodeSO(x,encWei,7))
carsCats$maxLoad <- names(encMlo)
carsSO$maxLoad <- sapply(cf$maxLoad,function(x) encodeSO(x,encMlo,7))
carsCats$displace <- names(encDsp)
carsSO$displace <- sapply(cf$displace,function(x) encodeSO(x,encDsp,7))
carsCats$maxPowKW <- names(encMkw)
carsSO$maxPowKW <- sapply(cf$maxPowKW,function(x) encodeSO(x,encMkw,7))
carsCats$maxPowKM <- names(encMkm)
carsSO$maxPowKM <- sapply(cf$maxPowKM,function(x) encodeSO(x,encMkm,7))
carsCats$rpm_maxPow <- names(encRMp)
carsSO$rpm_maxPow <- sapply(cf$rpm_maxPow,function(x) encodeSO(x,encRMp,7))
carsCats$maxTorque <- names(encMtq)
carsSO$maxTorque <- sapply(cf$maxTorque,function(x) encodeSO(x,encMtq,7))
carsCats$rpm_maxTor <- names(encRMt)
carsSO$rpm_maxTor <- sapply(cf$rpm_maxTor,function(x) encodeSO(x,encRMt,7))
carsCats$transmiss <- c(levTrans,NA)
carsSO$transmiss <- as.integer(factor(cf$transmiss,levels=levTrans))
carsCats$breaks <- c(levBreak,NA)
carsSO$breaks <- as.integer(factor(cf$breaks,levels=levBreak))
carsCats$minFuelCon <- names(encFue)
carsSO$minFuelCon <- sapply(cf$minFuelCon,function(x) encodeSO(x,encFue,7))
carsCats$accelTime <- names(encAcc)
carsSO$accelTime <- sapply(cf$accelTime,function(x) encodeSO(x,encAcc,7))
carsCats$maxSpeed <- names(encSpe)
carsSO$maxSpeed <- sapply(cf$maxSpeed,function(x) encodeSO(x,encSpe,7))
save(carsSO,file="./cars/cars.Rdata")
nCats <- as.integer(sapply(carsCats,length))
so <- emptySO(nCats)
# !!! problems with ' in car names; ' -> "'" in the file cars.nam
cn <- read.table("./cars/cars.nam",header=FALSE,sep=";")
long <- as.vector(cn[,2])
namedSO <- so
names(namedSO) <- c(names(carsCats),"num")
for(i in 1:nVar) names(namedSO[[i]]) <- carsCats[[i]]
save(nVar,nVarP,so,namedSO,numSO,long,carsCats,file="./cars/cars.meta")
SOs <- vector("list",numSO)
for(i in 1:numSO){
st <- so
for(j in 1:nVar) st[[j]][carsSO[[j]][[i]]] <- 1
st$num <- 1
names(SOs)[[i]] <- long[[i]]
SOs[[i]] <- st
}
save(nVar,nVarP,so,numSO,SOs,file="./cars/cars.so")
The encoded data are saved to ''cars.Rdata'' and further transformed and saved to ''cars.meta'' and ''cars.so''.
===== Clustering =====
From the file {{:pub:zip:specific.zip}} run the commands on ''analysisCars2.R''.
The cars data are stored on two files ''cars.so'' (symbolic objects) and ''cars.meta'' (metadata - names of units, variables and categories, ...).
After loading the data we run the leaders method ''leaderSO''. I started with 10 steps,
5 steps, 3 steps, 3 steps, ... The local minimum was attained in 22th step.
Then we apply the hierarchical clustering ''hclustSO'' to leaders and plot the corresponding dendrogram. Finally we list the units in cluster 10.
> # gDist - Cars data / Analysis
> # VB: 30.7.2010
> setwd("C:/Users/Batagelj/work/clamix/clamix.R")
> source("C:\\Users\\Batagelj\\work\\clamix\\clamix.R\\clamix2.R")
> load("./cars2/cars.so")
> load("./cars2/cars.meta")
> alpha <-rep(1/nVar,nVar)
> rez <- leaderSO(SOs,25)
Step 1
clust
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
81 156 16 106 30 72 164 71 36 5 14 62 18 81 14 48 63 5 19 30 40 29 120
24 25
52 17
[1] 0.3923260 0.4598792 0.3595308 0.3851367 0.3937892 0.3979975 0.3761325 0.4420161 0.4101287
[10] 0.3806951 0.3824818 0.3799591 0.3841496 0.4098006 0.3720880 0.4007372 0.3568077 0.3663876
[19] 0.3798095 0.4108323 0.3424131 0.3495385 0.3975587 0.3872848 0.3801581
[1] -0.3923260 -0.4598792 -0.3595308 -0.3851367 -0.3937892 -0.3979975 -0.3761325 -0.4420161
[9] -0.4101287 -0.3806951 -0.3824818 -0.3799591 -0.3841496 -0.4098006 -0.3720880 -0.4007372
[17] -0.3568077 -0.3663876 -0.3798095 -0.4108323 -0.3424131 -0.3495385 -0.3975587 -0.3872848
[25] -0.3801581
[1] 24.894921 55.357940 4.995569 32.930782 10.302790 23.652320 49.411368 22.531687 13.268613
[10] 1.741471 4.518207 19.784571 5.659036 28.469946 4.440819 15.499487 19.835808 1.677481
[19] 6.198282 10.350611 12.508263 9.155846 37.397783 17.595627 5.444684
[1] 437.6239
Times repeat = 10
Step 2
....
[1] 284.6798
Times repeat = 2
Step 22
clust
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
101 100 49 46 68 101 56 56 36 22 33 47 62 69 14 81 77 30 34 28 55 51 44
24 25
42 47
[1] 0.2283072 0.3878423 0.2664915 0.3595681 0.3520013 0.3283766 0.3229371 0.4070963 0.3272792
[10] 0.3230292 0.3791764 0.3377964 0.3153966 0.3232070 0.1942700 0.3897975 0.3347691 0.2461111
[19] 0.3148124 0.3168662 0.3368341 0.3432556 0.3338962 0.2575004 0.3015635
[1] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
[1] 14.287129 21.138846 8.134223 11.652174 16.217760 23.987814 8.723214 15.433379 7.371795
[10] 5.354895 6.217949 10.392799 12.830025 17.256410 1.741758 19.580247 17.799700 4.924359
[19] 7.065611 5.858516 11.089510 9.917044 9.193182 7.915751 10.595745
[1] 284.6798
Step 23
clust
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
101 100 49 46 68 101 56 56 36 22 33 47 62 69 14 81 77 30 34 28 55 51 44
24 25
42 47
[1] 0.2283072 0.3878423 0.2664915 0.3595681 0.3520013 0.3283766 0.3229371 0.4070963 0.3272792
[10] 0.3230292 0.3791764 0.3377964 0.3153966 0.3232070 0.1942700 0.3897975 0.3347691 0.2461111
[19] 0.3148124 0.3168662 0.3368341 0.3432556 0.3338962 0.2575004 0.3015635
[1] 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
[1] 14.287129 21.138846 8.134223 11.652174 16.217760 23.987814 8.723214 15.433379 7.371795
[10] 5.354895 6.217949 10.392799 12.830025 17.256410 1.741758 19.580247 17.799700 4.924359
[19] 7.065611 5.858516 11.089510 9.917044 9.193182 7.915751 10.595745
[1] 284.6798
Times repeat = 0
> save(rez,file="./cars2/cars25new.rez")
> hc <- hclustSO(rez$leaders)
> plot(hc,hang=-1)
> long[rez$clust==10]
[1] "HYUNDAI GALLOPER 2.5 T/D-XL" "JEEP WRANGLER 2.5 SPORT"
[3] "JEEP WRANGLER 4.0 SPORT" "JEEP WRANGLER 4.0 SPORT AUT"
[5] "JEEP WRANGLER 4.0 SAHARA" "JEEP WRANGLER 4.0 SAHARA AUT"
[7] "KIA SPORTAGE 2.0i MRi" "KIA SPORTAGE 2.0i MRDi"
[9] "KIA SPORTAGE 2.0 TD INTERCOOLER" "MITSUBISHI PAJERO 2.5 TD MT GL"
[11] "MITSUBISHI PAJERO 2.5 TD GLX" "MITSUBISHI PAJERO 2.8 TD MT GLS"
[13] "MITSUBISHI PAJERO 3.0 V6-24 MT GLS" "MITSUBISHI PAJERO 3.0 V6-24 MT GLS"
[15] "MITSUBISHI PAJERO 3.5 V6-24 MT GLS" "NISSAN TERRANO II SR H/5"
[17] "NISSAN TERRANO II SE H/T" "NISSAN TERRANO II TDi SR H/T"
[19] "NISSAN TERRANO II TDi SE H/T" "SSANGYONG KORANDO FAMILY RS"
[21] "TOYOTA RAV 4" "TOYOTA RAV 4 GX"
{{pics:notes:leadendnew.png}}