psuedocode for sequence clustering

sequence_pseudo.R

# normalize sequences
tx_sample <-
  tx_sample %>%
  mutate(
    order_date = order_date - first_date
  )
  
# create TraMineRextras object
tx_seq <- seqecreate(
  id = tx_sample$customer_id,
  timestamp = tx_sample$order_date,
  event = tx_sample$sale_type # this could be anything (email, return, channel, etc.)
)

# create distance matrix
seq_dist <- seqedist(
  seqe     = tx_seq,
  idcost   = rep(1, 3),
  vparam   = .1,
  interval = "previous"
)

# visualize distance matrix
seq_dist %>% cmdscale %>% plot

# hierarchical clustering
seq_clust <- agnes(
  x      = seq_dist,
  diss   = TRUE,
  method = 'ward'
)

# visualize some differences
seqedplot(tx_seq, type = "hazard", group = cutree(event_seq_normalized, 3))

# find discriminating subsequences
# not implemented yet—see http://traminer.unige.ch/doc/seqecmpgroup.html