Skip to content

Instantly share code, notes, and snippets.

@wey-gu
Created August 11, 2022 04:55
Show Gist options
  • Star 0 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save wey-gu/814cd311976d079e3b35f0590c04e174 to your computer and use it in GitHub Desktop.
Save wey-gu/814cd311976d079e3b35f0590c04e174 to your computer and use it in GitHub Desktop.
graphxr and nebulagraph demo 0: Graph tech based Fraud Detection
  • Fraud Detection https://siwei.io/fraud-detection-with-nebulagraph/
    • 思路是把我这个文章的方式藉由 GraphXR 的能力呈现出来,未来甚至可以用 iframe 来 embed 结果,像这个文章这种 https://distill.pub/2021/gnn-intro/
    • Graph Pattern based 方法,章节 1.2.1
      • 数据集 https://github.com/wey-gu/fraud-detection-datagen
      • 介绍风控专家提供的图模式,展示可视化
        • -- 针对一笔交易申请关联查询
          MATCH (n) WHERE id(n) == "200000010265"
          OPTIONAL MATCH p_shared_d=(n)-[:used_device]->(d)<-[:used_device]-(:applicant)-[:with_phone_num]->(pn:phone_num)<-[e:with_phone_num]-(:applicant)
          RETURN p_shared_d
      • GraphXR: 可以用更直观、优雅方式渲染这个模式
      • GraphXR: 可以抽取这个关系为直连的边:(n)-[:used_device]->(d)<-[:used_device]-(:applicant)-[:with_phone_num]->(pn:phone_num)<-[e:with_phone_num]-(:applicant)
      • GraphXR: 可以在抽取直连边之后做 View 内的 Louvain,进行团伙识别
      • NebulaGraph Algorithm,落地到生产 Pipeline 上,全图做等价的计算
    • 利用 Label Spread 进行标注扩充 章节 1.3
    • 利用带有图 feature 的 ML方法做分类章节 1.4
    • 利用 GNN 做 的方法,章节 1.5
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment