Practical Problem Solving - Java Data Mining: Strategy, Standard, and Practice - page 325

35. lResultSetCount

36. lStatement.executeQuery(lSQLCountQuery);

37. int lOtherTotalSize 0;

38. Map lOtherCategories new HashMap();

39. while (lResultSetCount.next()) {

40. int lCategorySize lResultSetCount.getInt(1);

41. String lCategory lResultSetCount.getString(2);

42. lOtherCategories.put(lCategory, new Integer(lCategorySize));

43. lOtherTotalSize lCategorySize;

44. }

To compute and compare frequencies in the two cases, we have
used hash maps to store the results and we have collected the global
number of customers in each case. For this, we scan through the two
maps and compare both categories and report this to the user. If we
go back to the situation in which we apply this to “PurchaseA,” we
could have received the following results for cluster 1:

Count PurchaseA
2345 1
45603 2
21342 3

And the following results for the population not in cluster 1:

Count PurchaseA
17546 1
31846 2
4275 3

These two result sets have been saved into two Java maps in order
to compute the profile in terms of relative frequencies and compute
the distance between these two distributions. The distance is based
on the average between the frequencies of the categories for each
population, as shown at line 58.

45. Set lCategoryNames lCategories.keySet();

46. Iterator lIter;

47. for (lIter lCategoryNames.iterator();

48. lIter.hasNext(); ) {

49. String lCategory (String)lIter.next();

50. double lOtherCategoryFreq 0.0;