##############################################################################################
#Function to convert lift from data mining reports to Relative Risk and Odds Ratio
#Description: calculate RR and OR from lift, support and confidence
#Instruction:   1) import the set of rules with lift, support and confidence (e.g. as csv)
#               2) run the function
#by Khanh Vu, University of Alberta, 2018
##############################################################################################

lift_RR <- function(rules, support, confidence, lift){
  k <- as.list(match.call())
  c = which(names(rules) %in% c(k$support, k$confidence, k$lift))
  support = eval(k$support, rules); confidence = eval(k$confidence, rules); lift = eval(k$lift, rules)
  prob_E = support / confidence 					            	#P(E)
  prob_O = confidence / lift                            				#P(O)
  RR = (1 - prob_E)*lift / (1 - prob_E*lift)
  OR = (RR - confidence) / (1 - confidence)							
  lift_O_negative = (1 - confidence) / (1 - prob_O) 		#Lift(O~|E)
  result <- data.frame(rules[-(c)], support, confidence, lift, RR, OR, prob_O, prob_E, lift_O_negative)
  return(result)
}    

#Example

#Create a set of rules with lift, support and confidence
rules = c("15 121 127 132 143 -> 4", "15 127 143 -> 4", "15 76 -> 4", "15 139 -> 4", "15 -> 4", "118 119 -> 4")
support = c(0.0196, 0.02, 0.0218, 0.0217, 0.0219, 0.0213)
cf =  c(0.115, 0.115, 0.113, 0.113, 0.112, 0.112)
gamma = c(1.295, 1.287, 1.266, 1.266, 1.263, 1.256)
Kingfisher = data.frame(rules, support, cf, gamma)

#call the function
example = lift_RR(rules = Kingfisher, support = support, confidence = cf, lift = gamma)
head(example)

#save result as .csv
write.table(example, "example.csv", sep=",", row.name = F)