Paper: arXiv 2403.00707
Authors: Adele Ravagnani, Fabrizio Lillo, Paola Deriu, Piero Mazzarisi, Francesca Medda, Antonio Russo
Abstract
Identification of market abuse is an extremely complicated activity that requires the analysis of large and complex datasets. We propose an unsupervised machine learning method for contextual anomaly detection, which allows to support market surveillance aimed at identifying potential insider trading activities. This method lies in the reconstruction-based paradigm and employs principal component analysis and autoencoders as dimensionality reduction techniques. The only input of this method is the trading position of each investor active on the asset for which we have a price sensitive event (PSE). After determining reconstruction errors related to the trading profiles, several conditions are imposed in order to identify investors whose behavior could be suspicious of insider trading related to the PSE. As a case study, we apply our method to investor resolved data of Italian stocks around takeover bids.
Complexity vs Empirical Score
- Math Complexity: 4.0/10
- Empirical Rigor: 6.0/10 (capped from 7.5: no out-of-sample test; see penalty flags)
- Quadrant: Street Traders — empirical and applied, lower math complexity (effective empirical rigor capped from 7.5)
Why this score: The paper employs standard dimensionality reduction techniques (PCA and autoencoders) with relatively light mathematical derivations, but demonstrates high empirical rigor by applying the method to a unique, large-scale, real-world dataset of investor trading records from Consob around specific price-sensitive events, yielding actionable, data-heavy insights for market surveillance.
Research Flowchart
flowchart TD
A["Research Goal: <br>Support Market Surveillance<br>for Insider Trading Detection"] --> B["Input: <br>Investor Trading Positions<br>around Price Sensitive Events PSEs"]
B --> C["Dimensionality Reduction<br>Techniques"]
C --> D["Principal Component Analysis PCA"]
C --> E["Autoencoders AE"]
D --> F["Compute Reconstruction Errors"]
E --> F
F --> G{"Apply Suspiciousness Conditions"}
G --> H["Outcomes: <br>Identification of Contextual Anomalies<br>Support for Market Abuse Detection"]
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 an unsupervised machine learning method for contextual anomaly detection using reconstruction-based PCA and autoencoders to support insider trading surveillance.
Source sentence
We propose an unsupervised machine learning method for contextual anomaly detection, which allows to support market surveillance aimed at identifying potential insider trading activities.
Assumes 6
- Investors’ positions are assumed to be zero at the initial time t0. data
Source sentence
In the definition of Equation 1, we assume that investors’ positions are null on t0.
- Investors with constant positions (no net trade) are discarded from the analysis. data
Source sentence
investors with constant positions i.e. investors who do not trade or are strict daily investors (i.e. the number of shares purchased and sold on each day are equal), are discarded.
- Normal observations have low reconstruction error, while anomalies have large reconstruction error. model
Source sentence
we expect that normal observations have low ϵi, while anomalies have large ϵi;
- Anomalous investors are substantially less numerous than normal investors. statistical
Source sentence
assuming they are substantially less numerous than investors with normal behavior.
Show 2 more
- The only input to the method is the trading position of each investor on the asset with a price sensitive event. data
Source sentence
The only input of this method is the trading position of each investor active on the asset for which we have a price sensitive event (PSE).
- Positions are computed in number of shares rather than monetary value. data
Source sentence
We also observe that positions are computed using the number of shares and not Euro.
Data
- Investor resolved data of Italian stocks around takeover bids other unknown
Source sentence
as a case study, we apply our method to investor resolved data of Italian stocks around takeover bids.
- Asset position dataset (shares per investor per day) other unknown
Source sentence
For each investor we extract from the database the asset position (in shares) at the end of each day.
Costs
Not applicable
Finds 4
- Among the first 500 ranked anomalous investors, 451 are also identified by the previous k‑means clustering method, showing high compatibility. performance
Source sentence
Of the first 500 ranked anomalous investors, 451 are also identified by the method based on k-means.
- The new dimensionality‑reduction approach can detect anomalous investors independently of the chosen starting point for position computation. performance
Source sentence
Another positive aspect of the new method based on a dimensionality reduction approach is its ability to detect given investors as anomalous, independently of the starting point chosen for the computation of her position
- The method based on PCA does not depend on knowledge of investors’ initial positions, unlike other insider detection methods. theoretical
Source sentence
Finally, it is worth noticing that, compared to other insider detection methods, the one based on PCA does not depend on the knowledge of the initial position of the investors.
- Autoencoders can capture investor types not detected by PCA, complementing the detection capability. performance
Source sentence
The ability of autoencoders to capture this type of investors which are not detected by PCA, was already shown in the study related to IMA.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 20 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.