The aim of bytd is to provide a toolkit for fitting the Beta-Geometric/Negative-Binomial Distribution (BG/NBD) model for CLV Modeling.
You can install the development version of bytd from GitHub with:
# install.packages("pak")
pak::pak("vitormarquesr/bytd")This is a basic example which shows you how to solve a common problem:
library(bytd)
# Simulating weekly RFM data for a base of 1300 customers
n <- 1300
# observation period of 80 weeks
T <- 80
# average purchase rate of twice per month
pars_gamma <- repar_Gamma(0.5, 0.3)
# average probability of droping out of 0.25.
pars_beta <- repar_Beta(0.25, 0.1)
rfm <- rbgnbd(n, T, pars_gamma[1], pars_gamma[2], pars_beta[1], pars_beta[2])
# fitting the marginalized loglikehood BG/NBD model throught a bayesian approach
m <- bgnbd(rfm, reff="marginal", approach="bayesian", cores=4)
m$fit
#> Inference for Stan model: bgnbd.
#> 4 chains, each with iter=2000; warmup=1000; thin=1;
#> post-warmup draws per chain=1000, total post-warmup draws=4000.
#>
#> mean se_mean sd 2.5% 25% 50% 75% 97.5%
#> r 2.52 0.00 0.15 2.24 2.42 2.52 2.62 2.85
#> alpha 4.89 0.01 0.37 4.23 4.64 4.88 5.12 5.69
#> a 4.37 0.02 0.81 3.17 3.80 4.25 4.83 6.37
#> b 13.32 0.09 2.94 8.99 11.27 12.83 14.94 20.33
#> lp__ -14775.11 0.04 1.40 -14778.71 -14775.79 -14774.78 -14774.07 -14773.38
#> n_eff Rhat
#> r 1612 1.00
#> alpha 1508 1.00
#> a 1065 1.01
#> b 1076 1.01
#> lp__ 1129 1.00
#>