マーケティングにおけるアトリビューションモデルのデモ

ChannelAttribution_demo.R

library(ChannelAttribution)
library(markovchain)
library(igraph)
library(ggplot2)

data(PathData)
H<-heuristic_models(Data,"path","total_conversions",var_value="total_conversion_value")
H
M<-markov_model(Data,"path","total_conversions",var_value="total_conversion_value",out_more=T)
M
transition_matrix<-M$transition_matrix
trans_conversion<-data.frame(channel_from="(conversion)",channel_to=unique(as.vector(transition_matrix$channel_to)),transition_probability=0)
trans_start<-data.frame(channel_from="(start)",channel_to=("start"),transition_probability=0)
final_transition<-rbind(rbind(transition_matrix,trans_conversion),trans_start)
transition_frame<-reshape(transition_matrix,direction = "wide", idvar="channel_from", timevar="channel_to")
transition_frame[is.na(transition_frame)]<-0
colnames(transition_frame)<-c("channel_from",as.vector(transition_frame$channel_from)[-1],"(start)")
rownames(transition_frame)<-transition_frame[,1]
transition_frame<-transition_frame[,-1]
finalmatrix<-as.matrix(transition_frame, dimnames = list(transition_frame$channel_from, colnames(transition_frame)[-1]))
final_obj<-new("markovchain",states=colnames(transition_frame),byrow=T,transitionMatrix=finalmatrix,name="ConversionPath")

plot(final_obj,edge.arrow.mode="<",vertex.size=10,vertex.color="skyblue",layout=layout.fruchterman.reingold,asp=0)

library(ChannelAttributionApp)
CAapp()