The specialized applications use all existing features of the basic Risk Framework structure and functionality, including system modules, models and the runtime core itself that interprets the models. System model sessions configure the work environment and are used to define basic objects, such as users, business hierarchy, settings, nomenclatures, catalogs etc. In addition, one or more models define the GUI and the specific business logic of the solution. These models are loaded into the Risk Framework environment to show the actual GUI, ensure interactions with users and perform calculations.
These solutions use the RFW data base to access common data, such as user data, nomenclatures, market data, etc. Specific data, that has not been saved in the RFW data base, can be entered using the GUI while a model session is created. Session data are saved and available for future work or reporting.
A common approach is to start the specialised solution according to its specific nature and then to define financial objects that describe the financial part of the solution. In this way, the total power of existing modules and functionalities of the platform is used:
The Compliance Check Solution in Risk Framework defines and evaluates restriction rules (for ex. limits, simple or structured conditions) that apply to selected portfolio subsets. The goal is to identify violations, produce warnings and and include them into reports. Depending on the portfolio objects rule sets represent:
In case of regulatory rule sets, the assignment of compliance rule sets can automatically be performed on all portfolios. General rule sets can be assigned to portfolio groups, while individual rule sets can be defined and applied on portfolio or sub-portfolio level.
The first step is to define the compliance structure of a portfolio, based on a tree hierarchy. Each tree node represents a portfolio subset (primary sub-portfolio), which is defined using Lists and Filters. Primary sub-portfolios are not necessarily exclusive. More complex sub-portfolios (sets) can be built from primary sub-portfolios, using set operations (Union, Intersection, Difference). Such sub-portfolio sets also belong to the compliance structure.
Rules are defined using GUI controls and can be edited and validated via formula editor, based on simple expression language. A rule can consist of data elements, arithmetic formulas, aggregation functions, comparison operators, and logical expressions. Aggregation functions (sum, max, min, average) are applied on sub portfolios and thus define more complex constraints.
Within a primary structure, rules can be assigned to tree nodes, sets of portfolios or single positions. The Compliance can be evaluated at any time - all constraints within a condition (rule) are checked in real time and warning and violation alerts are indicated. The distance to the restriction is colored, like a traffic light, in green, yellow and red and is represented in percentages.
Regulatory and non standard reports can be displayed on the screen or sent to supervision authorities for auditing.
For more information see Compliance Checks Workflow Concept.
Investment Consulting is a separate solution within the Risk Framework platform, which aims at supporting bank activities in order to attract investors. It offers investment proposals to clients who consider investing cash amounts or their own existing assets into a bank or an external depot. The bank officer can use the Investment Consulting solution to:
The risk class assessment of a client is based on his answers to a number of questions. The questions are assigned to specific question groups and each question group has a specific weight within the risk class calculation. The calculated risk class is mapped to a range ‘Low Limit-High Limit’, defined in the nomenclature of risk classes. Additional adjustments are made when the client’s risk class is closer to ‘Low Limit’ or to ‘High Limit’.
The abstract portfolio consists of bank product groups with appropriate risks and returns. All real and potential investments are evaluated for the entire investment period, provided the return is invested as well. Worst- and Best-case scenario projections of the investment are determined. Minimum, maximum, and risk free contributions on period basis are calculated.
Real instruments from a bank's instrument pool are mapped to product groups, so that the real investment portfolio is created. In this way, it is possible to observe the credit risk profile for the future investment period. A portfolio optimization is performed to achieve maximum return at minimum risk. The proposal is tailored to client's individual requirements and can be updated each time market conditions or client's financial states change, by creating a new investment consulting session.
The Investment Consulting solution can work along the Compliance Check solution to verify the validity against defined rule sets and alert the valuation of certain limits.
Investment proposals are saved into the database, displayed on the screen or printed out in different report formats (for ex. Crystal Reporter, QlikView, Excel, XML).
The Investment Consulting solution is controlled by the Settings Module, that isconcerned with nomenclatures (of products, market segments, loans, deposits), risk classes, model depots, and questionnaires. These elements can be defined in a manner.
The real estate and fund management solution is intended for assessment and management of investments in real estates. It is a separate software tool based on Risk Framework. The RFW structure modules and models, such as Catalogs, Nomenclatures, Instruments, Positions, Portfolios, etc., are used to create a working environment. Additional models are used to describe the real estate in Risk Framework as property or fund object, with specific features and valuation methods. Positions and portfolios of property objects, that are defined in Risk Framework, represent the basis of the investment analysis.
The work with properties in Risk Framework includes the following operations:
Models ‘Property Data’ and ‘Fund Data’ aim at identifying properties and funds, as well as define common property features, that asses property ratings, perform ratio analysis, list restrictions and register property contracts. The financial contracts of a property are represented in Risk Framework as positions. A real estate portfolio contains positions (property contracts) of one or several properties. Portfolio structuring allows the performing of Portfolio Evaluation, Cash Flow Analysis, and Interest Income Analysis, which represent the essence of real estate investment management. The solution for evaluation and management of funds includes both sides: the fund object side and the fund investor side.
The Time series analysis solution is an universal solution for analyses of historical time series, such as prices, foreign exchange rates, indexes, interest rates, interbanking rates, volatilities, etc. The tool is implemented as Java modules and can be invoked and used in Risk Framework and Risk Engine, as well as other software systems, to perform the following main tasks:
For more information see Mathematical Modelling Methods For Time Series.
Calculations in time series analysis are based on neuronal network approaches, such as self-organizing map (SOM) and prediction machine. The neuronal network is trained using unsupervised learning to produce the solution structure.
Clustering is the grouping of time series, such as bonds, shares, funds and other time developments, into clusters. In this way, series with similar historical behavior are grouped into the same cluster. Every cluster is then represented by its own synthetic time series (prototype), which enables the work with one series, instead of all real cluster series.
Inter and intra cluster statistics can be used to determine the optimal number of clusters, i.e. the most effective number of groups of minimal internal distance and maximal distance to each other. Clustering is used to reduce lare numbers of series and thus facilitate feasible time consuming operations, such ascalculations of huge correlation matrices, etc. Clustering can be used to reveal similar behaviors, for example, price negotiations between competitors, etc.
In most cases, the time series distribution for market factors is assumed to be normal. This does not correspond to the reality, though. Time series often expose skewed and flat tail distributions, which is connected to under-estimation of market risk for improbable large loses (flat tail losses).
For time series with non-normal distribution, it is important to identify the distribution type and its parameters. Numerical estimation of the distance between the empirical distribution and all other standard distribution types (for ex. Beta, Cauchy, Student, Weibull, etc.) allows users to choose the best fitting distribution type. This is important for the Copula Monte Carlo VaR simulation, which uses correlated non-normal distribution samples, instead of correlated normal distribution samples.
The goal of the multifactor modeling is to build a polynomial formula, which describes an unknown market instrument using instruments with known pricing models, based on time series best fit. After the formula has been created, it is periodically calibrated, using target instrument and explanatory instrument time series. After a target factor has been selected, explanatory factors must be chosen, either by automatic suggestion and/or manually. Explanatory factors are obtained from a cluster in which the target factor is classified. The generated formula can be used to develop a new type of instrument, that will have a pricing approach based on a set of known factors.
Basel III requires rating scales to be based on explicit CDS (Credit Default Swap) time series (CDS spread curves or indices, bond prices, share prices, etc.). Available time series are used to build and constitute a given number of rating degrees and to determine their boundaries, as well as to forecast the next probable rating that will correspond to a new series. Implied ratings and rating tendencies of an asset are determined based on its CDS Curve or CDS-Index (CDX) by attaching greater significance to the last value and using the weighting of series values (EWMA by Decay Factor).
By analysing given time series, the module can predict the behavior for a given time horizon, using the neural network. In addition to that, confidence bounds of predicted values and prediction quality statistics can be obtained. The learn effect is obtained by a sliding window, for example the period of 20 days, which is shifted along the history. The historical value within this window is used as a learning sequence and the values in the window are used to correct and enhance the learned behavior. A prediction occurs once the window reaches the end of the time series, where future values are pro-
The strategic liquidity management concerns the planning and management of future cash flows and long-term liquidity of a company and assesses possible, unexpected, and in many cases, unfavorable developments of business conditions and their impact to the company. Liquidity planning is based on the analysis and identification of known, suspected or potential outflows and evaluated alternative business strategies within the company, to ensure that adequate cash inflows are available that cover the outgoing payments where regulatory liquidity requirements are met.
The liquidity management modules offer effective technologies and procedures for monitoring, measuring and controlling of the liquidity positions:
The list of specialized software solutions can easily be expanded to cover various client requirements, specified as integrated application within the Risk Framework suit and the Risk Engine system. Our plan is to develop solutions for analyses and optimizations in the following areas: Ship Management, Fund Management, Management of products, Trading of products, Energy solutions , etc.