Organizations: College of Computer Science and Software Engineering, Shenzhen University Shenzhen, China · Alibaba Group Holding Limited Hangzhou, China · Guangdong Laboratory of Artificial Intelligence and Digital Economy (SZ) Shenzhen, China · Cyberspace Institute of Advanced Technology, Guangzhou University Hangzhou, china
Office automation (OA) systems play a crucial role in enterprise operations and management, with access control flow approval (ACFA) being a key component that manages the accessibility of various resources. However, traditional ACFA requires approval from the person in charge at each step, which consumes a significant amount of manpower and time. Its intelligence is a crucial issue that needs to be addressed urgently by all companies. In this paper, we propose a novel relational modeling-driven intelligent approval (RMIA) framework to automate ACFA. Specifically, our RMIA consists of two core modules: (1) The binary relation modeling module aims to characterize the coupling relation between applicants and approvers and provide reliable basic information for ACFA decision-making from a coarse-grained perspective. (2) The ternary relation modeling module utilizes specific resource information as its core, characterizing the complex relations between applicants, resources, and approvers, and thus provides fine-grained gain information for informed decision-making. Then, our RMIA effectively fuses these two kinds of information to form the final decision. Finally, extensive experiments are conducted on two product datasets and an online A/B test to verify the effectiveness of RMIA.
Figures & tables
Figure 1. An example of an access control flow approval (ACFA) process, where each data resource application is considered a step. The applicant initiates a permission application for a series of data resources due to specific business needs, and each step is reviewed and approved by the corresponding approver. Note that there are different permission types depending on the functional scope.
Figure 2. The architecture of our Relational Modeling-driven Intelligent Approval (RMIA) framework, which primarily includes a binary relation extractor for applicants and approvers, and a ternary relation extractor for applicants, resources, and approvers.
Figure 3. Example of a ternary relation extractor.
Datasets
Business-1
Business-2
Methods
AUC ↑
Precision ↑
Recall ↑
F1-score ↑
AUC ↑
Precision ↑
Recall ↑
F1-score ↑
LR
0.7685
0.8800
0.9600
0.9200
0.8622
0.9900
0.6900
0.8100
Random Forest
0.7969
0.8700
0.8700
0.9300
0.9232
0.9900
0.8800
0.9300
LightGBM
0.8024
0.8800
0.9800
0.9300
0.9308
0.9900
0.8800
0.9300
XGBoost
0.8052
0.8700
0.8600
0.9300
0.9267
0.9900
0.8700
0.9300
Ensemble
0.7905
0.8800
0.9800
0.9200
0.9209
0.9900
0.8500
0.9200
Table 1. Results on all datasets, where the best results are marked in bold. Note that ∗ indicates a significance level of p≤0.05 based on a two-sample t-test between our framework and the best baseline.
Figure 4. AUC results of ablation study on the Business-2 dataset.
Categories
Regular Employees
Internship Staff
Outsourced Employees
Ratio (%)
+1.93
+0.85
0.00
Table 2. The impact of applicants’ identity information on approval decisions.
Categories
Applicant’s Supervisor
Ordinary Resource Approvers
Ratio (%)
+2.86
0.00
Table 3. The impact of approver’s identity information on approval decisions.
Categories
C5
C4
C3
C2
C1
C0
Ratio (%)
+5.67
+4.81
+3.94
0.00
+2.31
+4.17
Table 4. The impact of affinity on approval decisions.
Categories
Business Resource Descriptions
General Resource Descriptions
Fuzzy Descriptions
Ratio (%)
+5.09
+3.74
0.00
Table 5. The impact of application text on approval decisions.
Figure 5. Overview of the online ACFA intelligent approval platform.
Figure 6. Overview of the number of permission requests processed and the model’s overall predicted approval rate each day within a month.
Metrics
AUC
Precision
Recall
F1-score
Base (LightGBM+BERT+Transformer)
0.0%
0.0%
0.0%
0.0%
RMIA
+4.90%
+10.34%
+8.92%
17.77%
Table 6. Gains obtained by our RMIA in online deployments.