英语论文网

1022LowHighTotal87.5%12.5%100.0%32246368LowHighTotal84.7%15.3%100.0%29353346LowHighTotal59.3%40.7%100.0%12318452076LowHighTotal80.9%19.1%100.0%1125265139015.5%85.5%100.0%106580686LowHighTotalFigure 6: Decision tree of the data set at time t in the same attribute and the same cut.Table 6: Number of changed rules in the same attribute and the same cut at time tþkType of change Number of changed rules Number of significantchanged rulesEmerging patterns 4 3 (degree of change >0.1)Unexpected changes 0 0 (degree of change >0.1)Added=Perished rules 0 0 (degree of change >0.001)Table 7: Significant emerging patterns (degree of change >0.1) in the same attribute and the same cut at time tþksupt(ri) suptþk(rj) aij1 If Sales visit¼Low, Reserved money¼Low, then Sales amount¼Low 0.57 0.19 0.66662 If Sales visit¼Low, Reserved money¼High, Payment¼Cash, thenSales amount¼Low0.25 0.12 0.52193 If Sales visit¼Low, Reserved money¼High, Payment¼Card, thenSales amount¼Low0.14 0.08 0.45994 If Sales visit¼High, then Sales amount¼High 0.33

Expert Systems, September 2005, Vol. 22, No. 4the same attribute and same cut of the tree, we built thedecision tree and rules of the data set at time t. Then, wecompared the two rule sets of time tþk and t to discover theemerging pattern, unexpected changes and added=perishedrules. Figures 5 and 6 show the decision trees of time tþkand time t respectively. The emerging patterns, unexpectedchanges and added=perished rules of the same attribute andthe same cut at time t are summarized in Table 6. Significantemerging patterns are summarized in Table 7.From changed rule (1) in Table 7, we can see the rapiddecrease of sales in the customer group who have a littlecyber money and visit the mall infrequently. In otherwords, these customers are likely to have disappeared in thenear future because of the high rate of decrease. Thereforean urgent market strategy is necessary to prevent the rapidsales decrease of these customers. With regard to unexpectedchanges and added=perished rules, we identifiedno significant changes.5.3. New decision tree at time t and time tþkThe decision trees are built with the data sets of time t andtime tþk independently using the SAS 8.0 enterprise minerprogram. And two rule sets are constructed along with thepathway of each decision tree. Then, we compared the rulesof time t and tþk and discovered the emerging patterns,unexpected changes and added=perished rules. We evaluatedthe changed rules to find what rules had changedmost. Figures 3 and 5 represent the new decision trees oftime t and time tþk respectively. Table 8 shows theemerging patterns, unexpected changes and added=perishedrules of the decision trees at time t and time tþk.Compared to the previous cases, new decision trees at eachtime generate more added rules and perished rules.Significant emerging patterns and added=perished rules attime t and time tþk are summarized in Tables 9 and 10.With regard to added=perished rules, we identified ninesignificant changes. From Table 10, we find that sales ofcustomers who are in their 40s are increased in the seconddata set from rule 5 and rule 8. This means that theimportance of customer groups who are in their 40s isincreased. Therefore additional services and products forsuch customers should be developed.6. ConclusionIn this pape本论文由英语论文网提供整理，提供论文代写，英语论文代写，代写论文，代写英语论文，代写留学生论文，代写英文论文，留学生论文代写相关核心关键词搜索。