Clustering and Visualization of Retail Market Baskets - Advanced Techniques in Knowledge Discovery and Data Mining - page 92

C Top product $ Lift Sec. product $ Lift Third product $ Lift

1 Bath gift packs 3.44 7.69 Hair growth m 0.90 9.73 Boutique island .81 2.61

2 Smoking cessati 0.15 34.73 tp canning item2.0418.74 Blood pressure 1.69 34.73

3 Vitamins other 3.56 12.57 tp coffee maker 1.46 10.90 Underpads hea 1.31 16.52

4 Games items 180 3.10 7.32 Facial moisturi .80 6.04 tp winejugite 1.25 8.01

5 Batt alkaline i 4.37 7.27 Appliances item3.65 11.99 Appliances appl 2.00 9.12

6 Christmaslight 8.11 12.22 Appliances hair 1.61 7.23 tp toaster/oven 0.67 4.03

7 Christmas food 3.42 7.35 Christmascards 1.99 6.19 Cold bronchial 1.91 12.02

8 Girl toys/dolls 4.13 12.51 Boy toys items 3.42 8.20 Everydaygirls 1.85 6.46

9 Christmasgiftw 12.51 12.99 Christmashome 1.24 3.92 Christmas food 0.97 2.07

10 Christmasgiftw 19.94 20.71 Christmaslight 5.63 8.49 Pers cd player 4.28 70.46

11 tp laundry soap 1.20 5.17 Facialcleanser 1.11 4.15 Hand&body thera 0.76 5.55

12 Film cameras it 1.64 5.20 Planners/calend0 94 5.02 Antacidh2 bloc 0.69 3.85

13 Tools/accessori .46 11.17 Binders items 2 .59 10.16 Drawing supplie 1.96 7.71

14 American greeti .42 5.34 Paperback items 2.69 11.04 Fragrances op 2.66 12.27

15 American greeti .56 6.72 Christmascards 0.45 2.12 Basket candy it .44 1.45

16 tp seasonalboo 10.78 15.49 American greeti 0.98 1.18 Valentineboxc 0.71 4.08

17 Vitaminseitem .76 6.79 Group stationer 1.0111.55 tp seasonalboo 0.99 1.42

18 Halloween bagc 2.11 6.06 Basket candy it 1.23 4.07 Cold cold items 1.17 4.24

19 Hairclrperman 12.00 16.76 American greeti1.11 1.34 Revlon cls face 0.83 3.07

20 Revlon cls face 7.0526.06 Hairclrperman 4.14 5.77 Headache ibupro 2.37 12.65

C Top product $ Lift Sec. product $ Lift Third product $ Lift

1 Actionitems 30 .26 15.13 tp video comedy 0.19 15.13 Family items 300 14 11.41

2 Smoking cessati 0.15 34.73 Blood pressure 1.69 34.73 Snacks/pnts nut 0.44 34.73

3 Underpads hea 1.31 16.52 Miscellaneous k 0.53 15.59 tp irons items .47 14.28

4 Acrylics/gels/w 0.19 11.22 tp exercise ite 0.15 11.20 Dental applianc0 81 9.50

5 Appliances item .65 11.99 Housewares peg 0.13 9.92 tp tarps items .22 9.58

6 Mult i ples p a cks 0. 17 13 . 87 Chr i st ma sl i ght 8 . 11 12 . 22 tv's i te m s 6 0. 44 8 . 32

44
8 Batt rechargeab .34 21.82 tp razors items .28 21.82 tp metalcookwa 0.39 12.77
9 tp furniture it 0.45 22.42 tp art&craft al .19 13.77 tp family plan, .15 13.76
10 Pers cd player 4.28 70.46 tp plumbing ite 1.71 56.24 Umbrellas adult 0.89 48.92
11 Catlitter scoo 0.10 8.70 Child acetamino .12 7.25 Pro treatment i 0.07 .78
12 Heaters items 8 0.16 12.91 Laverdiere ca 0.14 10.49 Ginseng items 4 0.20 6.10
13 Mop/broom lint .17 13.73 Halloween cards 0.30 12.39 Tools/accessori .46 11.17
14 Dentalrepairk 0.80 38.17 tp lawn seed it .44 35.88 tp telephones/a2.20 31.73
15 Gift boxes item 0.10 8.18 Hearing aidbat . 8 .25 American greeti5.56 6.72
16 Economydiapers 0.21 17.50 tp seasonalboo 10.78 15.49 Girls socks ite 0.16 12.20
17 tp wine 1.5lva 0.17 15.91 Group stationer 1.0111.55 Stereos items 2 0.13 10.61
18 tp med ointliq0.10 8.22 tp dinnerware i 0.32 7.70 tp bath towels 0.12 7.28
19 Hairclrperman 12.00 16.76 Covergirl imple 0.14 11.83 tp power tools 0.25 10.89
20 Revlon cls face 7.0526.06 Telephones cord 0.56 25.92 Ardell lashes i 0.59 21.87
Table 3.1. List of descriptive (top) and discriminative products (bottom) dominant
in each of the 20 value balanced clusters obtained from the drugstore data (see also
Figure 3.4(f)). For each item the average amount of money spent in this cluster
and the corresponding lift are given.

7 Sleep ai ds i te m 0. 31 14 . 61 K a v a k a v ai te m s . 51 14 . 21 tp beer super p 0. 14 12 .

times more on blood pressure-related items, peanuts, and snacks. Do these
customers lead an unhealthy lifestyle and are they eager to change? Cluster
C 15 , which can be seen to be a highly compact cluster of Christmas shoppers
characterized by greeting-card and candy purchases. Note that

had
an extra constraint that clusters be of comparable value. This may force a
larger natural cluster to split, as may be the case causing the similar clusters
C 9 and

Opossum

C 10 . Both are Christmas-gift shoppers (Table 3.1(top)), cluster C 9 are

the moderate spenders, and cluster C 10

is much smaller with equal revenue contribution (Fig. 3.4(f)). Our hunch is
reinforced by looking at Fig. 3.4(f).

C 10 are the big spenders, as cluster