Credit Risk modelling, model validation, management, and capital requirements determination

The Analytics Boutique provides a set of best of breed tools for Credit management, modelling and determination of capital requirements (Basel`s Pillar 2 and Internal Models under Solvency II). They can be used for the following purposes:

  • Credit MachineLearning: create models for rating, scoring, PD, EAD, LGD and Other
  • SMachineLearning Validation: Generate validation statistics and reports for the above models
  • Loss Forecasting: develop stress testing models based on macroeconomic factors
  • Credit Portfolio Risk Analyser: determine capital requirements under Basel’s Pillar I and II, including capital allocations for risk-adjusted performance management
  • Acceleris: Loan origination, rating and pricing based on all the above models
  • Integrated Financial Resource Management: Stress testing and ICAAP

Our credit risk solutions cover the entire value chain ranging from the development of credit risk models to the application in day-to-day management of the business:

Credit MachineLearning
Credit MachnieLearning’s (CML) objective is to improve precision, integrity, efficiency and governance in credit predictive modelling by use of cutting-edge methodologies (AI) and making them available to every organisation. We enable your team to develop ‘best-in-class’ models:
  • Permits to create all Credit Risk models: rating, scoring, PD, LGD, EAD, CCF and others
  • No programming skills required: user-friendly interface and integrated dataflows completely eliminates need for coding; self explanatory interface; re-do and un-do functions similar to MS Office
  • Maximal Prediction Accuracy: Champion-Multi challenger fitting method: simultaneous fitting and ranking of all models (under any metric) to identify the most predictive model; AI applied to model fitting, feature binning, sample selection, best fit model identification and more; exhaustive coverage of modelling methods and ‘goodness- of-fit’ metrics
  • Strong model governance: Full audit trail of all modelling steps; options and user control permits to enforce bank`s modelling policy; User control permits to flexibly define roles and activities: modeller, validator, administrator, ...; modelling database stores all developed (unlimited) models incl. development data sample/history classified by BU, purpose, version, data scientists and other
  • Highest efficiency in model building & validation: Exhaustive library of modelling methods incl. required steps and phases; central database storing ‘work in progress’ to allow later continuation of work
  • Integrity guaranteed in model validation: Audit Trail, reporting and modelling history protected with digital hash (block chain technology); permits to validate any report using the digital hash, allowing a strong control over large teams of modellers; modelling database keeps all models together with their development sample, modelling history (all the steps from raw development sample to final model) and audit trail, permitting a deep and efficient validation
  • Exhaustive reporting by the press of the button: Complete model report with the push of a bottom outlining all results incl. graphs, plots and descriptive stats; permits a deep and efficient validation, by keeping development sample and modelling steps in the central database
Credit Loss Forecasting
Following the same user friendliness and governance principles as in Credit Machine Learning, Credit Loss Forecasting focuses on building time series analysis for stress testing and loss forecasting:
  • Creates stress testing models (PD, LGD...) for credit loss forecasting under such scenarios, based on time series analysis, ARIMA, ARIMAX, Block bootstrapping, Lasso, Ridge and others
  • Champion-challenger for identifying most robust and stable models
  • Projection tool permitting to estimate credit risk variable values under such scenario
  • Storage of times series
  • Full model governance: audit trail, user control, workflow, etc.
  • Extensive reporting by the push of a button
  • No programming skills required
Credit MachineLearning Validation
  • Extensive validation metrics for Machine Learning and traditional predictive credit risk models (PD, transition matrices, CCF, LGD, IFRS9, ...):
  • Full reporting by the push of a button
  • Strong model governance: audit trail, user control and other
Credit Portfolio Risk Analyser - CPRAxx
CPRAxx is a powerful Monte Carlo simulation based credit portfolio modelling solution that can be used for various kinds of credit concentration risk analyses (in particular credit economic capital) that feed into other credit risk management processes such as risk-adjusted performance measurement/pricing, ICAAP, credit stress testing, risk appetite and limit setting
  • Underlying methodology uses well established CreditMetrics approach with key enhancements like PD/LGD correlations or importance sampling
  • User-friendly web interface enables fast preparation of model runs
  • Availability of multiple capital allocation approaches with detailed output regarding the corresponding simulation results
  • Parallel computing capabilities allow for usage of multi-CPU server environments
  • Simulation mode (mark-to-market vs. default/no-default) can be assigned at an individual instrument level
  • New deals can be assessed for pricing purposes without having to re-run the entire portfolio
  • Open source code to further customise tool to clients’ requirements
Acceleris Loan Management
Acceleris provides a fully controlled, high precision and efficient process for loan origination, rating and pricing. It first calculates an accurate client rating grade and corresponding PD. Then, the loan pricing is calculated based on the PD and incorporating LGD, EAD, financial and operational costs and detailed transaction characteristics
  • End-to-end web platform for loan origination, rating, PD calculation, pricing and post approval facility management
  • Loan rating grade and PD
  • Loan pricing: spread and reference rate
  • Transaction profitability: RAROC, expected loss, interest rates paid, present value from economic and accounting views, transaction cost structure and more
  • Audit trail of client and loan life cycle status evolution
  • Model validation data for the backtesting of PDs, evaluation of loan pricing precision, rating/scoring models validation and recalibration and more
  • Loan portfolio’s ratings, PDs, LGDs and EADs for capital requirement calculations
  • Strategic reports defined based on two sources:
  • Portfolio composition: reports on any information captured or calculated during the rating and pricing including ratings, statuses, expected loss, profitability metrics (NPV, average RAROC and other) and so on
  • Log of actions : reports on the activity and/or activity quality by user/group. For instance, the number of loans transacted by user/group, or the average PD assigned by user/group, or the number of rating iterations by user
Integrated Financial Resource Management (IFRM)
It is a web-based platform that allows the efficient evaluation of multiple business and macroeconomic scenarios (including stress scenarios) by forward-looking projecting P&L, BS, capital requirements & availability and liquidity analysis. It provides a consistent and integrated view of financial planning, strategy evaluation, risk measurement and stress testing:
  • P&L: Pre-provision net income, provisions etc.
  • Impairment (IFRS 9 expected credit losses, including breakdown of stages 1-3)
  • Capital requirements based on RWAs for credit, market and operational risks
  • Balance sheets with required drilled down on asset classes
  • Funding structure
  • Liquidity analysis (LCR)
  • Dividend policy
  • Views by business unit, business lines, country, legal entity, etc.
  • Ratio analysis: performance, risk appetite...
  • Accepts any number of user-defined external risk factors: macroeconomic (GDP, unemployment, ...), financial/market risk (IR, FX, ...), business risk (volumes, spreads, operating costs...), hedging structure, customer behaviour (behavioural at maturity, prepayments, product transition across front, middle and back books...)


We always strive for excellence in what we do