Urban association rules: uncovering linked trips for shopping behavior

📝 Abstract

In this article, we introduce the method of urban association rules and its uses for extracting frequently appearing combinations of stores that are visited together to characterize shoppers’ behaviors. The Apriori algorithm is used to extract the association rules (i.e., if -> result) from customer transaction datasets in a market-basket analysis. An application to our large-scale and anonymized bank card transaction dataset enables us to output linked trips for shopping all over the city: the method enables us to predict the other shops most likely to be visited by a customer given a particular shop that was already visited as an input. In addition, our methodology can consider all transaction activities conducted by customers for a whole city in addition to the location of stores dispersed in the city. This approach enables us to uncover not only simple linked trips such as transition movements between stores but also the edge weight for each linked trip in the specific district. Thus, the proposed methodology can complement conventional research methods. Enhancing understanding of people’s shopping behaviors could be useful for city authorities and urban practitioners for effective urban management. The results also help individual retailers to rearrange their services by accommodating the needs of their customers’ habits to enhance their shopping experience.

💡 Analysis

In this article, we introduce the method of urban association rules and its uses for extracting frequently appearing combinations of stores that are visited together to characterize shoppers’ behaviors. The Apriori algorithm is used to extract the association rules (i.e., if -> result) from customer transaction datasets in a market-basket analysis. An application to our large-scale and anonymized bank card transaction dataset enables us to output linked trips for shopping all over the city: the method enables us to predict the other shops most likely to be visited by a customer given a particular shop that was already visited as an input. In addition, our methodology can consider all transaction activities conducted by customers for a whole city in addition to the location of stores dispersed in the city. This approach enables us to uncover not only simple linked trips such as transition movements between stores but also the edge weight for each linked trip in the specific district. Thus, the proposed methodology can complement conventional research methods. Enhancing understanding of people’s shopping behaviors could be useful for city authorities and urban practitioners for effective urban management. The results also help individual retailers to rearrange their services by accommodating the needs of their customers’ habits to enhance their shopping experience.

📄 Content

Urban  association  rules:  uncovering  linked  trips  for  shopping   behavior  

  Yuji  Yoshimuraa,b,  Stanislav  Sobolevskya,c,  Juan  N  Bautista  Hobina,  Carlo   Rattia,  Josep  Blatb  

  a  SENSEable  City  Laboratory,  Massachusetts  Institute  of  Technology,  77   Massachusetts  Avenue,  Cambridge,  MA  02139,  USA;  

  b  Department  of  Information  and  Communication  Technologies,  Universitat   Pompeu  Fabra,  Roc  Boronat,  138,  Tanger  Building  08018  Barcelona,  Spain;  

  c  Center  for  Urban  Science  and  Progress,  New  York  University,  1  MetroTech  Center,   19th  Floor,  Brooklyn,  NY  11201,  USA;  

  Abstract.  In  this  article,  we  introduce  the  method  of  urban  association  rules  and   its  uses  for  extracting  frequently  appearing  combinations  of  stores  that  are  visited   together   to   characterize   shoppers´   behaviors.   The   Apriori   algorithm   is   used   to   extract  the  association  rules  (i.e.,  if  -­‐>  result)  from  customer  transaction  datasets   in   a   market-­‐basket   analysis.   An   application   to   our   large-­‐scale   and   anonymized   bank   card   transaction   dataset   enables   us   to   output   linked   trips   for   shopping   all   over  the  city:  the  method  enables  us  to  predict  the  other  shops  most  likely  to  be   visited  by  a  customer  given  a  particular  shop  that  was  already  visited  as  an  input.   In  addition,  our  methodology  can  consider  all  transaction  activities  conducted  by   customers  for  a  whole  city  in  addition  to  the  location  of  stores  dispersed  in  the  city.   This  approach  enables  us  to  uncover  not  only  simple  linked  trips  such  as  transition   movements   between   stores   but   also   the   edge   weight   for   each   linked   trip   in   the   specific   district.   Thus,   the   proposed   methodology   can   complement   conventional   research  methods.  Enhancing  understanding  of  people’s  shopping  behaviors  could   be   useful   for   city   authorities   and   urban   practitioners   for   effective   urban   management.  The  results  also  help  individual  retailers  to  rearrange  their  services   by  accommodating  the  needs  of  their  customers’  habits  to  enhance  their  shopping   experience.  

  Keywords:  shopping  behaviors,  association  rule,  transaction  data,  Barcelona  

  1.  Introduction  

  In   this   paper,   we   explore   the   applicability   of   association   rules   (Agrawal   et   al.,   1993)   for   extraction   of   combinations   of   visited   stores   for   the   analysis   of   linked   trips   for   shopping   behavior.   The   Apriori   algorithm   (Agrawal   &   Srikant,   1994),   which   is   widely   used   and   a   rather   simple   yet   robust   method   for   market-­‐basket   analysis,   is   applied   to   our   anonymized   large-­‐scale   transaction   dataset.   This   algorithm   was   originally   designed   to   extract   combinations   of   other   items   most   likely  to  be  purchased  by  a  customer  given  a  particular  item  that  is  already  in  his   or  her  basket  as  an  input.  Instead  of  the  purchased  items  in  a  single  store,  we  try  to   extract  the  links  between  the  stores  in  which  people  make  transactions  before  or   after   visiting   some   focal   shops,   considering   all   transactions   conducted   in   stores   dispersed  throughout  the  city.  Thus,  we  can  uncover  the  edge  weight  for  each  trip   as   the   transition   probability   between   stores   by   comparing   it   with   the   general   pattern  of  all  other  shoppers’  behaviors  in  the  given  district.  

  An   anonymized   bankcard   transaction   dataset   provides   us   with   longer-­‐term   evidence  that  people  perform  transactions  among  stores  as  a  digital  footprint  or   “data   exhaust“   (Mayer-­‐Schönberger   &   Cukier,   2013,   p113)

This content is AI-processed based on ArXiv data.