Estimates parameters for the (M)BG/CNBD-k model via Maximum Likelihood Estimation.
mbgcnbd.EstimateParameters(
cal.cbs,
k = NULL,
par.start = c(1, 3, 1, 3),
max.param.value = 10000,
trace = 0
)bgcnbd.EstimateParameters(
cal.cbs,
k = NULL,
par.start = c(1, 3, 1, 3),
max.param.value = 10000,
trace = 0
)
mbgnbd.EstimateParameters(
cal.cbs,
par.start = c(1, 3, 1, 3),
max.param.value = 10000,
trace = 0
)
Calibration period customer-by-sufficient-statistic (CBS)
data.frame. It must contain a row for each customer, and columns x
for frequency, t.x for recency , T.cal for the total time
observed, as well as the sum over logarithmic intertransaction times
litt, in case that k is not provided. A correct format can be
easily generated based on the complete event log of a customer cohort with
elog2cbs.
Integer-valued degree of regularity for Erlang-k distributed
interpurchase times. By default this k is not provided, and a grid
search from 1 to 12 is performed in order to determine the best-fitting
k. The grid search is stopped early, if the log-likelihood does not
increase anymore when increasing k beyond 4.
Initial (M)BG/CNBD-k parameters. A vector with r,
alpha, a and b in that order.
Upper bound on parameters.
If larger than 0, then the parameter values are is printed every
trace-step of the maximum likelihood estimation search.
A vector of estimated parameters.
(M)BG/CNBD-k: Reutterer, T., Platzer, M., & Schroeder, N. (2020). Leveraging purchase regularity for predicting customer behavior the easy way. International Journal of Research in Marketing. 10.1016/j.ijresmar.2020.09.002
Batislam, E. P., Denizel, M., & Filiztekin, A. (2007). Empirical validation and comparison of models for customer base analysis. International Journal of Research in Marketing, 24(3), 201-209. 10.1016/j.ijresmar.2006.12.005
# NOT RUN {
data("groceryElog")
cbs <- elog2cbs(groceryElog)
(params <- mbgcnbd.EstimateParameters(cbs))
# }
Run the code above in your browser using DataLab