Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Empowering recommender systems using automatically generated Knowledge Graphs and Reinforcement Learning

Ghanshyam Verma, Shovon Sengupta, Simon Simanta, Huan Chen, Janos A. Perge, Devishree Pillai, John P. McCrae, Paul Buitelaar

TL;DR

This work tackles interpretable personalized recommendations in finance by leveraging automatically generated knowledge graphs (KGs). It formalizes the task as a deterministic Markov Decision Process over a KG and develops two KG-driven approaches: a reinforcement learning method with Path Directed Reasoning (PDR) and a KG-driven XGBoost baseline that uses KG embeddings (TransE and Tucker) with post-hoc explanations. Automatic KG generation from structured and unstructured data yields variants such as cKG, uKG_DP, and uKG_CN, with embeddings informed by TransE and Tucker decompositions; explanations are produced via SHAP and ELI5. On a real-world subscriber-educational article dataset, the KG-RL model achieves the best MAP@K and overall predictive performance, while KG embeddings improve XGBoost-based recommendations, demonstrating both accuracy gains and enhanced interpretability for customer relationship management.

Abstract

Personalized recommender systems play a crucial role in direct marketing, particularly in financial services, where delivering relevant content can enhance customer engagement and promote informed decision-making. This study explores interpretable knowledge graph (KG)-based recommender systems by proposing two distinct approaches for personalized article recommendations within a multinational financial services firm. The first approach leverages Reinforcement Learning (RL) to traverse a KG constructed from both structured (tabular) and unstructured (textual) data, enabling interpretability through Path Directed Reasoning (PDR). The second approach employs the XGBoost algorithm, with post-hoc explainability techniques such as SHAP and ELI5 to enhance transparency. By integrating machine learning with automatically generated KGs, our methods not only improve recommendation accuracy but also provide interpretable insights, facilitating more informed decision-making in customer relationship management.

Empowering recommender systems using automatically generated Knowledge Graphs and Reinforcement Learning

TL;DR

This work tackles interpretable personalized recommendations in finance by leveraging automatically generated knowledge graphs (KGs). It formalizes the task as a deterministic Markov Decision Process over a KG and develops two KG-driven approaches: a reinforcement learning method with Path Directed Reasoning (PDR) and a KG-driven XGBoost baseline that uses KG embeddings (TransE and Tucker) with post-hoc explanations. Automatic KG generation from structured and unstructured data yields variants such as cKG, uKG_DP, and uKG_CN, with embeddings informed by TransE and Tucker decompositions; explanations are produced via SHAP and ELI5. On a real-world subscriber-educational article dataset, the KG-RL model achieves the best MAP@K and overall predictive performance, while KG embeddings improve XGBoost-based recommendations, demonstrating both accuracy gains and enhanced interpretability for customer relationship management.

Abstract

Personalized recommender systems play a crucial role in direct marketing, particularly in financial services, where delivering relevant content can enhance customer engagement and promote informed decision-making. This study explores interpretable knowledge graph (KG)-based recommender systems by proposing two distinct approaches for personalized article recommendations within a multinational financial services firm. The first approach leverages Reinforcement Learning (RL) to traverse a KG constructed from both structured (tabular) and unstructured (textual) data, enabling interpretability through Path Directed Reasoning (PDR). The second approach employs the XGBoost algorithm, with post-hoc explainability techniques such as SHAP and ELI5 to enhance transparency. By integrating machine learning with automatically generated KGs, our methods not only improve recommendation accuracy but also provide interpretable insights, facilitating more informed decision-making in customer relationship management.
Paper Structure (17 sections, 6 equations, 3 figures, 1 table, 1 algorithm)

This paper contains 17 sections, 6 equations, 3 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. RELATED WORK
  3. Collaborative Filtering
  4. Knowledge Graph-driven Recommender Systems
  5. Reinforcement Learning based Recommender Systems
  6. METHODOLOGY
  7. KG-Driven Reinforcement Learning based Recommender System
  8. KG-Driven XGBoost based Recommender System
  9. Experimental Setup
  10. Automatic KG Generation
  11. ConceptNet-based approach
  12. Dependency Parsing-based approach
  13. KG creation using both structured and unstructured data (cKG)
  14. Knowledge Graph Embeddings
  15. Data sets
  16. ...and 2 more sections

Figures (3)

  • Figure 1: Explaining the recommendations of RL-based approach using the path in the KG that leads to the recommendation.
  • Figure 2: Explaining the recommendations of KG-XGBoost [uKG_CN] model using SHAP.
  • Figure 3: Explaining the recommendations of KG-XGBoost [uKG_CN] model using ELI5.