“Dimension Reduction: Feature Subset” Method for Selecting the Best Index Combination in Reputation Evaluation of Crowdsourcing Participants

Table 1

Weight of index variables after dimension reduction using ReliefF.

Rank	Index variable	Weight
1	x 1 positive feedback rate	0.054
2	x 2 number of punishments in three months	0.037
3	x 3 refund rate in three months	0.029
4	x 4 refund rate in this month	0.027
5	x 5 credibility frozen after reporting	0.023
6	x 6 transaction activity	0.023
7	x 7 work attitude	0.015
8	x 8 task completion quality	0.015
9	x 9 work speed	0.014
10	x 10 years of opening	0.014
11	x 11 comprehensive scoring	0.012
12	x 12 number of employer nonrecommend	0.008
13	x 13 margin deposit amount	0.008
14	x 14 number of negative feedback	0.007
15	x 15 growth scoring	0.006
16	x 16 number of medium feedback	0.0055
17	x 17 number of punishments	0.037
18	x 18 employer repurchase rate	0.001
19	x 19 number of refunds in this month	0.001
20	x 20 number of refunds in three months	0.0003
21	x 21 city	0.00001
22	x 22 number of employer recommendation	−0.002
23	x 23 number of positive feedback	−0.0029
24	x 24 number of employers did not choose	−0.003
25	x 25 transaction amount in recent three months	−0.0003
26	x 26 number of employers served	−0.0037
27	x 27 transaction times of three months	−0.0004
28	x 28 store type	−0.0004