Paper: arXiv 2306.05987
Authors: Ruihua Ruan, Emmanuel Bacry, Jean-François Muzy
Abstract
Thanks to the access to the labeled orders on the CAC40 data from Euronext, we are able to analyze agents’ behaviors in the market based on their placed orders. In this study, we construct a self-supervised learning model using triplet loss to effectively learn the representation of agent market orders. By acquiring this learned representation, various downstream tasks become feasible. In this work, we utilize the K-means clustering algorithm on the learned representation vectors of agent orders to identify distinct behavior types within each cluster.
Complexity vs Empirical Score
- Math Complexity: 7.0/10
- Empirical Rigor: 6.5/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper employs advanced mathematical concepts like contrastive learning with triplet loss and vector space embeddings, which indicates high math complexity, while its use of labeled LOB data, specific clustering algorithms (K-means), and detailed data preprocessing suggests solid empirical implementation.
Research Flowchart
flowchart TD A["Research Goal<br>Represent Liquidity Taker Behavior"] --> B["Data Input<br>Labeled Orders CAC40 Euronext"] B --> C["Self-Supervised Learning<br>Triplet Loss Architecture"] C --> D["Computation<br>Learned Order Representation Vectors"] D --> E["Downstream Task<br>K-means Clustering"] E --> F["Outcome<br>Identified Distinct Agent Behavior Types"] F --> G["Conclusion<br>Effective Behavior Representation & Analysis"]
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 a self‑supervised contrastive learning model with triplet loss to learn representations of agent market orders and uses K‑means clustering on these representations to identify distinct liquidity‑taker behavior types.
Source sentence
In this study, we construct a self-supervised learning model using triplet loss to effectively learn the representation of agent market orders.
Assumes 1
- Individual agents maintain consistent behavior over short periods, while different agents exhibit distinct behaviors. model
Source sentence
We make the assumption that individual agents maintain consistent behavior over short periods, while different agents exhibit distinct behaviors.
Data
- CAC40 index future data order-book proprietary
Source sentence
Our study uses the Cac40 index future data provided by Euronext Paris, spanning from January 2016 to February 2017.
Costs
Not applicable
Finds 2
- K‑means clustering on the learned order‑representation vectors partitions the data into seven distinct clusters representing different liquidity‑taker behaviors. statistical
Source sentence
This clustering cut the set to seven clusters.
- The feature set Basic+M+QS (including time, quantity, side, limit‑to‑trade modification, best bid/ask prices, and best‑level queue sizes) is sufficient for the task, while adding deeper queue levels does not significantly improve performance. statistical
Source sentence
we conclude that these eight features listed in Table 1 are sufficient for our task.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 10 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.