Building and Applying Knowledge Graphs for Financial Asset Management with AI and Big Data

Why a Knowledge Graph for Finance? The financial asset‑management sector suffers from information and cognition asymmetry; a knowledge‑graph system helps investors overcome these gaps by providing structured, searchable, and up‑to‑date intelligence.

Intelligent Investment‑Research Process Data from heterogeneous sources are collected, transformed into indicators via a data‑middle‑platform, and combined with AI (NLP, machine learning) to generate actionable research intelligence.

System Architecture The platform consists of three layers: a data layer (massive multi‑source acquisition), a middle‑platform layer (data, knowledge, and algorithm services), and an application layer (investment scenarios such as portfolio simulation and risk control).

Data Ingestion Platform A cloud‑native, Kafka‑driven collection system built on Kubernetes handles TB‑scale daily data, supports rapid addition of new sources, and ensures real‑time stability across thousands of heterogeneous feeds.

AI‑Powered Web Extraction An AI‑driven web‑page parser uses tag‑embedding and a three‑layer feed‑forward classifier to filter irrelevant HTML nodes, then classifies retained nodes with a Hierarchy‑TextCNN + memory‑block model, achieving >97% tag‑filter accuracy and >95% content extraction accuracy.

Knowledge Matching & Extraction A hierarchical TextCNN model (later enhanced with relevance‑search) maps raw indicators to financial concepts; FinBERT‑based short‑text classification and summarisation extract facts from reports and news, while a PDF‑table extractor converts financial tables into structured data.

Entity Linking A transformer + CRF model identifies entities, and a FinBERT twin‑network performs disambiguation and ranking, achieving an F1 > 0.95 for entity extraction.

Knowledge Fusion Multiple facts about the same entity are merged by evaluating timeliness, authority, richness, and conflict, ensuring the graph contains concise, high‑quality knowledge.

Typical Application The integrated system powers an intelligent research assistant, exemplified by a 5G industry knowledge graph that combines market data, news, and sentiment to support investment decisions.