an experiment in GIST for R

test.r

# select a particlar car and day to examine
x.journey.number <- 4
x.day <- "Monday"

# Big look to process all the records
#for (x.journey.number in 1:length(l.journeys)) {
#} # end big loop


df.x <- read.csv(paste(x.results.dir,"pattern_",l.journeys[x.journey.number]
                       ,sep = ""), stringsAsFactors = FALSE)

# find the max cluster number in the data frame
x.max.cluster <- max(c(max(df.x$from),max(df.x$to)))

# initialize a matrix to calculate frequency and probability
a.x <- array(rep(0, 
                 (x.max.cluster)*(x.max.cluster)), 
             dim=c(x.max.cluster,x.max.cluster))

# calculate the journeys, and add a nominal amount to allow for some uncertainty 
# when there is 0 journeys (since we are not certain there are NEVER any journeys 
# in between location that in our sample reported as 0)
for (j in 1:x.max.cluster) {
  for (k in 1:x.max.cluster) {
    a.x[j,k] <- df.x[df.x$from == j &
                             df.x$to == k &
                             df.x$day == x.day,"journeys"] + 0.1
  }
}

# add some row and column names
rownames(a.x) <- rownames(a.x, do.NULL = FALSE, prefix = "From.")
colnames(a.x) <- colnames(a.x, do.NULL = FALSE, prefix = "To.")

# add the marginal sums
a.x <- rbind(a.x,colSums(a.x))
a.x <- cbind(a.x,rowSums(a.x))

# great a marginal probability estimate, based on normalized observed frequency
b.x <- round(a.x/a.x[x.max.cluster+1,x.max.cluster+1],3)