mit wojtekw @ RECONCILE --- !!! Z poważnymi błędami!!! ERROR

gistfile1.r

#wczytanie danych dotyczacych ocen stron przez mturkow
d_eval<-read.csv("/Users/andi/Desktop/RECONCILE/DANE_evaluations.csv", header=TRUE, sep=";")



library(matrixStats)
library(R.methodsS3)



#przygotowanie danych dla Troi
#z NA
troia_input<-subset(d_eval,select=c(person_id,document_id,documentevaluation_credibility))
write.table(troia_input, file="/Users/andi/Desktop/RECONCILE/troia_input.csv",row.names=FALSE, col.names=FALSE,sep=" ")

#bez NA
troia_input_without_na <- troia_input[!is.na(troia_input$documentevaluation_credibility),]
write.table(troia_input_without_na, file="/Users/andi/Desktop/RECONCILE/troia_input_without_na.csv",row.names=FALSE, col.names=FALSE, sep=" ")

#z 0 zamiast NA
troia_input_with_zeros <- troia_input
troia_input_with_zeros[is.na(troia_input_with_zeros)]<-0
write.table(troia_input_with_zeros, file="/Users/andi/Desktop/RECONCILE/troia_input_with_zeros.csv",row.names=FALSE, col.names=FALSE, sep=" ")



#proste statystyki, wykresy
#biblioteki plyr uzyjemy do szybkiego wygenerowania podsumowan 
library(plyr)

#zwroci tabele zawierajaca w pierwszej kolumnie liczbe ile-krotnie strona zostala oceniona, a w drugiej ile stron z calego zbioru dotyczyla tylokrotna ocena.
#wariant wliczajacy NA
count(count(d_eval,"document_id"),"freq")
#plot(count(count(d_eval,"document_id"),"freq"))
b<-barplot(count(count(d_eval,"document_id"),"freq")$freq.1,main="Distribution of number of ratings per page (including 0 - \"don\'t know\")",names.arg=count(count(d_eval,"document_id"),"freq")$freq)
mtext(side = 1, at = b, text = paste("#:",count(count(d_eval,"document_id"),"freq")$freq.1), line = 3)

#wariant pomijajacy NA
count(count(troia_input_without_na,"document_id"),"freq")
#plot(count(count(troia_input_without_na,"document_id"),"freq"))
b<-barplot(count(count(troia_input_without_na,"document_id"),"freq")$freq.1,main="Distribution of number of ratings per page(without 0 - \"don\'t know\")",names.arg=count(count(troia_input_without_na,"document_id"),"freq")$freq)
mtext(side = 1, at = b, text = paste("#:",count(count(troia_input_without_na,"document_id"),"freq")$freq.1), line = 3)



temp<-aggregate(temp,by=list(temp$document_id), FUN=mean, na.rm=TRUE)

srednia<-function(lista) {
# print(lista)
sum1=0
sum2=0
for (i in 1:length(lista)){
	sum1=sum1+reated(lista[i])*notna(lista[i])
	sum2=sum2+reated(lista[i])
}
if (sum2!=0){
as.numeric(sum1/sum2)
}
else if(sum1==0){
as.numeric(0)
}
}



wariancja<-function(lista) {
sum1=0
sum2=0
srednia=srednia(lista)
for (i in 1:length(lista)){
	sum1=sum1+reated(lista[i])*(notna(lista[i])-srednia)^2
	sum2=sum2+reated(lista[i])
}
if (sum2!=0){
as.numeric(sum1/sum2)
}
else if(sum1==0){
as.numeric(0)
}
}




reated <- function(arg){
#if(is.na(arg)) 0 else 1
if(is.na(arg)){
	as.numeric(0)
}
else{
	as.numeric(1)
	}
}
notna <- function(arg){
if(is.na(arg)) 0 else arg
}

# co najmniej 2 oceny=5 -> strona trafia do klasy HC (Highly Credibly)
# co najmniej 2 oceny=4 -> strona trafia do klasy N (Neutral)
# co najmniej 2 oceny<4 -> strona trafia do klasy HNC (Highly Not Credible)
adamw_classifier<-function(lista){
#zamiana NA na 0
for (i in 1:length(lista)){ lista[i]=notna(lista[i]) }
#stworzenie data frameu z czestosciami wystapien poszczegolnych ocen dla strony
df<-data.frame(table(lista))
if ((5 %in% df$lista) && (df[df$lista==5,]$Freq >= 2)) {		
			#print("HC was assigned")
			return (3)#(as.character("HC"))
	}
else if ((4 %in% df$lista) && (df[df$lista==4,]$Freq >= 2)){
					#print("N was assigned")
					return (2)#(as.character("N"))
	}
else if (sum(df[as.numeric(levels(df$lista)[df$lista]) < 4,]$Freq) >=2){
			#print("HNC was assigned")
			return (1)#(as.character("HNC"))
	}
else{
	#print("UNDIFINED was assigned")
	return (0)#(as.character("UNDIFINED"))
	}
}




tmp<-troia_input
length(unique(tmp$document_id))
dane<-ddply(tmp,.(document_id),summarise,SREDNIA=srednia(documentevaluation_credibility),WARIANCJA=wariancja(documentevaluation_credibility),CLASS_ADAMW=adamw_classifier(documentevaluation_credibility))

kmeans_data<-matrix(c(dane$SREDNIA, dane$WARIANCJA),ncol=2)
kmeans_result<-kmeans(kmeans_data,sum(unique(dane$CLASS_ADAMW)))
plot(y=jitter(dane$WARIANCJA,amount=0.2),x=jitter(dane$SREDNIA,amount=0.2))
plot(y=jitter(dane$WARIANCJA,amount=0.2),x=jitter(dane$SREDNIA,amount=0.2),col=dane$CLASS_ADAMW)
plot(y=jitter(dane$WARIANCJA,amount=0.2),x=jitter(dane$SREDNIA,amount=0.2),col=kmeans_result$cluster)
points(kmeans_result$centers,col=6,pch=10,cex=3)