Paper: arXiv 2311.00964
Authors: Chengyao Wen, Yin Lou
Abstract
Rules are widely used in Fintech institutions to make fraud prevention decisions, since rules are highly interpretable thanks to their intuitive if-then structure. In practice, a two-stage framework of fraud prevention decision rule set mining is usually employed in large Fintech institutions; Stage 1 generates a potentially large pool of rules and Stage 2 aims to produce a refined rule subset according to some criteria (typically based on precision and recall). This paper focuses on improving the flexibility and efficacy of this two-stage framework, and is concerned with finding high-quality rule subsets in a bi-objective space (such as precision and recall). To this end, we first introduce a novel algorithm called SpectralRules that directly generates a compact pool of rules in Stage 1 with high diversity. We empirically find such diversity improves the quality of the final rule subset. In addition, we introduce an intermediate stage between Stage 1 and 2 that adopts the concept of Pareto optimality and aims to find a set of non-dominated rule subsets, which constitutes a Pareto front. This intermediate stage greatly simplifies the selection criteria and increases the flexibility of Stage 2. For this intermediate stage, we propose a heuristic-based framework called PORS and we identify that the core of PORS is the problem of solution selection on the front (SSF). We provide a systematic categorization of the SSF problem and a thorough empirical evaluation of various SSF methods on both public and proprietary datasets. On two real application scenarios within Alipay, we demonstrate the advantages of our proposed methodology over existing work.
Complexity vs Empirical Score
- Math Complexity: 7.0/10
- Empirical Rigor: 8.5/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper employs advanced multi-objective optimization theory (Pareto fronts, hypervolume indicators, heuristic frameworks) and mathematical categorization of SSF problems, while demonstrating strong empirical rigor through extensive experiments on public and proprietary datasets, including real-world Alipay applications with concrete metrics.
Research Flowchart
flowchart TD A["Research Goal: Improve two-stage fraud rule mining<br>in Fintech (Precision & Recall trade-off)"] --> B["Data Inputs:<br>Public & Alipay Datasets"] B --> C["Stage 1: SpectralRules Algorithm<br>Generates compact, diverse rule pool"] C --> D["Intermediate Stage: PORS Framework<br>Applies Pareto Optimality to find Non-Dominated Rule Subsets"] D --> E["Core Process: Solution Selection on Front (SSF)<br>Categorization & Heuristic Optimization"] E --> F["Stage 2: Pareto Front Analysis<br>Flexible selection of optimal rule subsets"] F --> G["Key Findings:<br>1. SpectralRules improves subset quality<br>2. PORS simplifies selection criteria<br>3. Validated via Alipay applications"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes SpectralRules, a novel sequential covering algorithm for diverse rule generation, and PORS, a heuristic framework for Pareto-optimal rule subsets.
Source sentence
We propose a novel variant of the sequential covering algorithm called SpectralRules that directly generates a compact pool of rules with high diversity, and demonstrate on both public and proprietary datasets that SpectralRules improves HV over existing tree-based approaches [16] for Stage 1.
Assumes
No explicit assumptions found in the text.
Data
No dataset named in the text.
Costs
Not applicable
Finds 5
- SpectralRules improves hypervolume (HV) over existing tree-based approaches for Stage 1. performance
Source sentence
We propose a novel variant of the sequential covering algorithm called SpectralRules ... that
- PORS with the hvc-ss SSF method outperforms CRSL on most public and proprietary datasets. performance
Source sentence
PORS with hvc-ss outperforms CRSL on most cases.
- There is a positive correlation between HV (intermediate metric) and recall (final objective). statistical
Source sentence
we observe that there is a positive correlation between HV (as an intermediate evaluation metric) and recall (as a final objective).
- PORS algorithm converges quickly, whereas NSGA‑II is less efficient. performance
Source sentence
PORS algorithm quickly converges while NSGA-II is less efficient.
- High diversity of rules in Stage 1 leads to higher HV. performance
Source sentence
Combined with results in Section 5.2, we conclude that high diversity in Stage 1
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 40 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (50)
- Github repo (2024) — link
- Interpretable decision sets: A joint framework for description and prediction (2016)
- C4. 5: programs for machine learning (1993)
- Classification and regression trees (1984)
- Random forests (2001)
- Interpretable Machine Learning (2020)
- Fast effective rule induction (1995)
- The CN2 induction algorithm (1989)
and 42 more
- Mining association rules between sets of items in large databases (1993)
- Diverse Rule Sets (2020)
- Information Retrieval (1979)
- A bagging SVM to learn from positive and unlabeled examples (2014)
- Building text classifiers using positive and unlabeled examples (2003)
- PEBL: positive example based learning for web page classification using SVM (2002)
- Partially supervised classification of text documents (2002)
- Towards Consumer Loan Fraud Detection: Graph Neural Networks with Role-Constrained Conditional Random Field (2021)
- Online Credit Payment Fraud Detection via Structure-Aware Hierarchical Recurrent Neural Network (2021)
- Intention-aware heterogeneous graph attention networks for fraud transactions detection (2021)
- Pick and Choose: A GNN-based Imbalanced Learning Approach for Fraud Detection (2021)
- Spatio-temporal attention-based neural network for credit card fraud detection (2020)
- Interleaved sequence rnns for fraud detection (2020)
- Financial defaulter detection on online credit payment via multi-view attributed heterogeneous information network (2020)
- LightGBM: A highly efficient gradient boosting decision tree (2017)
- ‘Could You Describe the Reason for the Transfer?’ A Reinforcement Learning Based Voice-Enabled Bot Protecting Customers from Financial Frauds (2021)
- Uncovering insurance fraud conspiracy with network learning (2019)
- Flowscope: Spotting money laundering based on graphs (2020)
- Nextgen AML: Distributed deep learning based language technologies to augment anti money laundering investigation (2018)
- Learning with positive and unlabeled examples using weighted logistic regression (2003)
- Bagging predictors (1996)
- An Adaptive Framework for Confidence-constraint Rule Set Learning Algorithm in Large Dataset (2022)
- The Hypervolume Indicator: Problems and Algorithms (2020) — link
- Benchmarking subset selection from large candidate solution sets in evolutionary multi-objective optimization (2022) — link
- Quality evaluation of solution sets in multiobjective optimisation: A survey (2019)
- Evolutionary algorithms for multiobjective optimization: Methods and applications (1999)
- Equispaced Pareto front construction for constrained bi-objective optimization (2013)
- Fast Greedy Subset Selection From Large Candidate Solution Sets in Evolutionary Multiobjective Optimization (2021)
- Comparison of multiobjective evolutionary algorithms: Empirical results (2000)
- Clustering-based subset selection in evolutionary multiobjective optimization (2021)
- A fast and elitist multiobjective genetic algorithm: NSGA-II (2002)
- MOEA/D: A multiobjective evolutionary algorithm based on decomposition (2007)
- Benchmarking multi-and many-objective evolutionary algorithms under two optimization scenarios (2017)
- Many-objective evolutionary algorithms: A survey (2015)
- SPEA2: Improving the strength Pareto evolutionary algorithm (2001)
- SMS-EMOA: Multiobjective selection based on dominated hypervolume (2007)
- HypE: An algorithm for fast hypervolume-based many-objective optimization (2011)
- SPEA2+: Improving the Performance of the Strength Pareto Evolutionary Algorithm 2 (2004)
- Solution Subset Selection for Final Decision Making in Evolutionary Multi-Objective Optimization (2020)
- A Combined Genetic Adaptive Search (GeneAS) for Engineering Design (1996)
- SMPSO: A new PSO-based metaheuristic for multi-objective optimization (2009)
- Fanglue: An Interactive System for Decision Rule Crafting (2023)
0 of 50 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.