Table 5 AP comparison of all models on journal entry data under the i.i.d. setting. The presentation style is consistent with that of TableĀ 3.

From: Anomaly detection in double-entry bookkeeping data by federated learning system with non-model sharing approach

Ī»

4

8

AP

APall ↑

APglobal ↑

APlocal ↑

APall ↑

APglobal ↑

APlocal ↑

IA

0.496

(0.064)

0.913

(0.097)

0.196

(0.070)

0.496

(0.064)

0.913

(0.097)

0.196

(0.070)

FedAvg

0.585​

(0.087)​

0.857​

(0.138)​

0.326​

(0.083)​

0.614​

(0.035)​

0.813

(0.164)​

0.375​

(0.074)​

FedProx

0.471

(0.056)

0.990

(0.029)

0.149

(0.065)

0.491

(0.107)

0.985

(0.046)

0.191

(0.152)

DC (PCA)

0.513

(0.077)

0.990

(0.021)

0.198

(0.109)

0.512

(0.037)

1.000

(0.000)

0.206

(0.063)

DC (RP)

0.458 (0.050)

1.000 (0.000)

0.150 (0.070)

0.467 (0.030)

1.000 (0.000)

0.141 (0.049)

CA

0.686​

(0.060)​

0.958​

(0.075)​

0.440​

(0.099)​

0.686​

(0.060)​

0.958​

(0.075)​

0.440​

(0.099)​