Summary
"Consider a scenario in a financial market where investors privately share important insider information within their social circles. Although financial authorities could monitor investors' transactions, it remains challenging to identify which investors have received and are exploiting insider information. I term this scenario a hidden cascade problem, in which the true states of agents cannot be directly observed. This particular form of cascade modelling, involving the inference of agent states from indirect observations, has received limited attention in the existing literature.
This proposal, ""HiddenTipChains"", introduces approaches based on Topological Data Analysis, Machine Learning, and probabilistic modeling to reveal hidden information transfer through investors’ social connections and to detect suspicious trading based on insider information. This supports European Commission goals for financial stability and investor protection. In a broader context, the methods will enable us to analyse hidden cascade models when direct observations of agents' true states are not directly observable. I expect that in future research, these approaches can find application in addressing hidden cascade challenges across various fields, such as epidemiology, climate science, and information security.
The empirical part of the project is based on world-wide unique and exceptionally extensive datasets, providing us with full access to investors’ full market-wide trading history on all the securities they traded. Additionally, we have observable social connections among insider investors, allowing us to track how news tips about future events propagate within this network. The approached developed in this project will be verified with synthetic data with extensive Monte-Carlo experiments. The project contributes to the Computer Science literature by developing techniques for hidden cascade problems and within the Quantitative Finance domain for identifying market abuse.
"
This proposal, ""HiddenTipChains"", introduces approaches based on Topological Data Analysis, Machine Learning, and probabilistic modeling to reveal hidden information transfer through investors’ social connections and to detect suspicious trading based on insider information. This supports European Commission goals for financial stability and investor protection. In a broader context, the methods will enable us to analyse hidden cascade models when direct observations of agents' true states are not directly observable. I expect that in future research, these approaches can find application in addressing hidden cascade challenges across various fields, such as epidemiology, climate science, and information security.
The empirical part of the project is based on world-wide unique and exceptionally extensive datasets, providing us with full access to investors’ full market-wide trading history on all the securities they traded. Additionally, we have observable social connections among insider investors, allowing us to track how news tips about future events propagate within this network. The approached developed in this project will be verified with synthetic data with extensive Monte-Carlo experiments. The project contributes to the Computer Science literature by developing techniques for hidden cascade problems and within the Quantitative Finance domain for identifying market abuse.
"
Unfold all
/
Fold all
More information & hyperlinks
| Web resources: | https://cordis.europa.eu/project/id/101150609 |
| Start date: | 01-07-2024 |
| End date: | 30-06-2026 |
| Total budget - Public funding: | - 199 694,00 Euro |
Cordis data
Original description
"Consider a scenario in a financial market where investors privately share important insider information within their social circles. Although financial authorities could monitor investors' transactions, it remains challenging to identify which investors have received and are exploiting insider information. I term this scenario a hidden cascade problem, in which the true states of agents cannot be directly observed. This particular form of cascade modelling, involving the inference of agent states from indirect observations, has received limited attention in the existing literature.This proposal, ""HiddenTipChains"", introduces approaches based on Topological Data Analysis, Machine Learning, and probabilistic modeling to reveal hidden information transfer through investors’ social connections and to detect suspicious trading based on insider information. This supports European Commission goals for financial stability and investor protection. In a broader context, the methods will enable us to analyse hidden cascade models when direct observations of agents' true states are not directly observable. I expect that in future research, these approaches can find application in addressing hidden cascade challenges across various fields, such as epidemiology, climate science, and information security.
The empirical part of the project is based on world-wide unique and exceptionally extensive datasets, providing us with full access to investors’ full market-wide trading history on all the securities they traded. Additionally, we have observable social connections among insider investors, allowing us to track how news tips about future events propagate within this network. The approached developed in this project will be verified with synthetic data with extensive Monte-Carlo experiments. The project contributes to the Computer Science literature by developing techniques for hidden cascade problems and within the Quantitative Finance domain for identifying market abuse.
"
Status
SIGNEDCall topic
HORIZON-MSCA-2023-PF-01-01Update Date
22-11-2024
Images
No images available.
Geographical location(s)
Search
