Demonstration of analytic hierarchy process applied to district heat load forecasting

district-heat-ahp-demo.r

## District heating AHP example
##
## This example demonstrates how the analytic hierarchy process can be
## used to select a preferred technique for forecasting the thermal
## load of a district heating system.
##
## James Keirstead (j.keirstead@imperial.ac.uk)
## Presented 25 June 2013 at ISIE Conference, Ulsan, South Korea
##

## Load required libraries
library(devtools)
library(plyr)

## The AHP code is available from Github
gist_url <- "https://gist.github.com/jkeirstead/5827295"
source_gist(gist_url)

## Define the criteria and methods to be evaluated
criteria <- c("Accuracy", "Speed", "Theory")
methods <- c("Linear regression", "Time series", "Neural networks")

## Calculate the weights for each method by criteria
criteria_weights <- data.frame()
for (crit in criteria) {
  w <- AHP(methods, crit)
  criteria_weights <- rbind(criteria_weights,
                            data.frame(criteria=crit, methods=methods, weights=w$weight))
}

## Calculate the weights of the criteria to the goal
overall_weights <- AHP(criteria, "Thermal load prediction")

## Combine everything and run it
tmp <- merge(criteria_weights,
             data.frame(criteria=criteria, cr.weight=overall_weights$weight))
tmp2 <- transform(tmp, total=weights*cr.weight)
result <- ddply(tmp2, .(methods), summarize, score=sum(total))

## Tidy the result and print
result <- arrange(result, desc(score))
print(result)

## Final plot
library(ggplot2)
gg <- ggplot(result, aes(x=methods, y=score)) +
  geom_bar(stat="identity") +
  theme_bw() +
  labs(x="", y="AHP Score")
ggsave("district-heat-ahp.pdf", gg, width=8, height=6)