Big Data for Risk Management
Big Data Risk Management Solution for Banks
Investment banks need up-to-date information about their current and historical risk exposures. It is necessary for compliance with bank regulations as well as demanded by clients and shareholders.
Traders, risk management and compliance departments need not only to know current Value at Risk per trade or position but also up-to-date aggregated VaR (Value at Risk) per trade book, per floor, per client and even for whole bank.
We have gathered data from risk calculation systems such as Atlas, Giraffe or SAS as well as market data and current trades. We have bypassed complex ETL (Extract Transform Load) processes and converted VaR, market data and trades into common format in Hadoop platform. We have calculated aggregated VaR where necessary and gave interested parties immediate access to current and historical values of VaR, trades and market data.
Compliance and risk management department can see not only market data as been seen by trader at the moment of trade but also VaR and aggregated VaR with different granularities available at the moment to the trader. This is enabled by Big Data system that stores all VaR and market data together with trade data and enables quick retrieval.
Value at Risk VaR
Investment banks need up-to-date information about their risk exposure. Risk measures are calculated from market fluctuations during certain time period and existing positions. The most important measure of risk is Value At Risk (VaR). If exposure to certain risk is too high traders try to hedge their positions by buying futures, selling certain shares and so on.
Staled Aggregated Risk Data in Current Systems
Risk measures per trade or position are calculated base on market data in trading systems, aggregated risks are calculated in risk management systems. These systems rely on daily imports of trades and market data. This leads to incomplete and stale picture of aggregated risks. Risk management teams and traders need to see up-to-date and complete picture of risks in order to make right decisions.
Non-Available Historical Risk Data at the Moment of the Trade
When trade turns sour it is important to investigate what went wrong. We have seen many situation where risk data and aggregated risk data were available only for whole day of trading. These values were not the same as seen by trader at the moment of the trade. Risk management system must be able to display historical aggregated risks as well as drill down through various aggregation levels such as books to the specific positions.
Big Data Approach
Big Data technologies enable us to store complete dataset for each trade. This dataset include trade information, market data and risk data available to the trader at the moment of each trade. We are able to quickly retrieve given data via simple web interface.
We have utilized existing Risk Calculation systems at the bank to calculate VaR and aggregated VaR. These systems often do not return aggregated VaR in time because lengthy imports of market data. We have decided to calculate missing VaR for different time periods from data that is currently available. These values are continuously replaced by more precise values from Risk Calculation systems and new calculations.
We have bypassed complex, lengthy, fragile and even manual ETL (Extract Transform Load) process by Big Data ETL and integration.
Integration into Current Environment of the Bank
There are currently multiple risk management, trading and market risk systems that offer output of measures such as Delta, Gamma, Vega, Rho, Theta and VaR in different formats, timescales and different levels of aggregation. We have used capabilities of these systems to export data into CSV (Comma Separated Values) format to file system. This capability is useful because we do not need to access their databases directly, we do not need to know internal data structures.