At oneof applications of the data mining,analyzingthe transaction (transaction) data of the POS data and the EC sight, it is the search data analysis which discovers the combination "of the commodity which is bought together".
The transaction data of 1 time by 1 customer is called the market basket data, the data mining is done this the transaction database which was totaled in week and the month unit as a source. Especially it makes effective with the volume sales retail industry, but it is possible also to apply to the transaction data of the credit card and the utilization record etc. of the communication system company.
Techniqueof associationanalysis is utilized generally, the result [ bread and butter ] -> [ milk ], is outputin the formof the association kind of rule which is said.
"The customer who buys the diaper buys also the beer simultaneously", that you say,the case "of the diaper and the beer " is famous. In addition, "8 tenths of person who purchases the latex paint at the Sunday carpenter store purchase the roller simultaneously" "8 tenths of person who purchases ?????????? at the food store, bottle entering ?????? is purchased simultaneously", "4 tenths of investor which has stock index fund have installed growth type investment trust fund in the portfolio", and so on the case is reported.
The analysis result refers, when layout inside the store and examining the plan of stack and the display, the selection and the commodity stock etc. of the articles for special sale ? sale campaign commodity.
For example, it understands that if as for loss leader A and the customer who purchases daily necessaries B, the probability which purchases quality item C is high, the case which prepares the set commodity of ABC, it can get near with the question ? callous which increases C stock, separating conversely, arranging, the excursion inside the store of the customer to urge, it sells at a special price A and B you can think of the arrangement of each one. In addition it is thought that C is recommended to the customer who does not buy A and B.
In addition due to the issue of the occurrence time data and the member card etc. of the time zone and day of the week and season etc. the customer including buying past record and the customer attribute data etc. which specify, by the fact that the analysis object (being called the virtual item) it makes, "it comes to weekday evening and buying pattern of the customer who the store is done" and "after buying item A, as for the customer whom it adds purchases item B, the classified by time that/the time-series data mining and the like becomes possible 80% purchases item C".
This way by the fact that measure of another commodity promotion and the like to the specific commodity buyer is taken, with convenience of the customer and the improvement of degree of satisfaction,you aim towardthe sale improvement and the like by the cross cell.
The association analysis which is utilized with market basket analysis, when the number (and the number of items which form association rule) of items increases, increases the number of combinations which it should consider explosively. In addition association analysis, when the transaction quantity of each item of the analysis object data is same, is a feature that the best result is easy to be produced.
Because of that, the analysis data is not to utilize the unit for stock management and sale management that way, for example as, "00 beer 350ml cans" "beer" or "alcoholic drinks", it is good to rearrange item classification according to purpose. In additionthe commodity whichcan be sold with ABC analysis and the like, it assumes that the hand that is effective, it makes the commodity which cannot be sold clear, increasing the taxonomic level of the commodity which cannot be sold, it collects widely. |
