Financial Analysis Using Dynamic Knowledge Structure


eCognition technology brings a new, dynamic approach to the problems of Financial Analysis. eCognition is all about making knowledge active, allowing much more knowledge about what needs to be analysed and simulated to be embedded in the financial model, and allowing the model itself to produce knowledge in a form which can be embedded in operational systems, allowing the circle of

analysis of inputs -> operational decisions -> analysis of results

to be closed. Because it works with knowledge, all of the essential knowledge handling techniques are present - visibility, extensibility, dynamic structure modification, management of consistency over time and across many contributors,.

Attributes of the Knowledge Approach

Rapid Implementation

With Tupai’s technology the implementation of an analytic system is not delayed by a long preparation phase of cleaning the data, mining it, and analyzing the results. The same components that are used in the model are also used to read the data, clean it if necessary, and convert it into useful information. The model can then be used to control what knowledge needs to be updated and when. Rapid implementation and updating ensures that all system components are consistent with what is happening now.

Ability to Merge and Integrate Disparate Forms of Knowledge

Tupai’s technology combines "concrete" declarative or analytical knowledge, with "fuzzy" and experiential knowledge. While some knowledge is analytic or hard (e.g. the timing of a new product release or the structure of sales incentives), much of the knowledge in the model is experiential and comes from data about return on investments in various markets, competitor activity, changing needs of populations as their demographics alter, and basic economic changes.

Only when all this knowledge is brought together in a unified way – regardless of knowledge type - can the relationship between the manufacturer and a particular customer be successfully managed. The challenge for a DFA system is to merge the different forms of knowledge about risk and be sufficiently up to date so the users of the system can make high level strategic decisions based on valid and unbiased views of what is occurring.

Simulation and Optimization Capabilities Assist Decision Support

The model provides simulation and optimization to support decision-making. All the available knowledge about the assets and liabilities can be taken into account in the simulation. This allows studying the consequences of a particular decision strategy, and updating the decisionmaking within the model using Machine Learning - the same Machine Learning that was used to introduce experiential knowledge into the model.

Rapid and Ongoing Response to Change

One thing is certain and that is that change will be ever present. Tupai’s active model responds to changes in the incoming data by changing its own structure, guaranteeing that it maintains its usefulness and does not become out of date shortly after the initial setup. Traditional approaches, which use an amalgam of spreadsheet and program modules, quickly become outdated as both the operation of the organisation and its external environment evolve. The snapshot approach has a further weakness, in that parts of the model, which have captured the essence of the operation at different times, may produce inconsistent results because they no longer fit together perfectly.

Explicit Handling of Uncertainty

Traditional models either use rigid analytic formulae to capture risk, or use stochastic models. Tupai's technology allows close integration of hypothetical and stochastic risk, with the ability to then manipulate the probabilities that arise from this integration. The technology enables the direct representation of uncertainty in the models, using distributions and relations that respond directly to other effects.

Time Series Analysis

Time Series Analysis is an integrated component of Tupai’s knowledge handling, allowing predictions to be based on all the other knowledge available, and those predictions fed back into the overall model. In this way, the seasonal and other factors that would mask small but important differences can be removed.

Precise Representation of the Business Process



Speed Versus Flexibility

The model allows all of the analysis to be handled in the network for maximum visibility and flexibility, and where it can be updated in seconds, or moved into script operators using an interpreter, where the instructions can be tested with all of the surrounding network information flows live, or moved into compiled operators for maximum speed.

Combining Analytic and Experiential Knowledge


Operators and Messaging

Search and Dataflow

Distributions and Relations

To handle situations where the analysis would be too complex to include in a simulation model, multi-dimensional relations can be used to store experiential knowledge. For example, future interest rates can be based on a mix of both historical experience and analysis taking into account effects like the inertia of the economy, long term reversion to a mean, volatility increasing with the deviation from the mean and the time constants of the business cycle. Realistic interest rate scenarios for simulation can be derived from this source, as well as allowing the simulation to have its own means of predicting forward scenarios to be used in choosing strategies for portfolio management.

Structural Backtrack

When the model runs, the objects in its portfolio are destroyed and new ones created. The model uses structural backtrack as the fastest way to revert to its starting state. This same backtrack can be used to step back from a simulation failure and explore strategies that will allow the model to get past the pinch point. The successful strategies can then be stored in relations and used when similar conditions are detected in future runs. The same strategy relations can also be embedded in operational decision support, ensuring the simulation model is a close approximation to the way the organisation is managed.

Working with the System

Building the Model

The model can be built and extended using graphical object or textual approaches, or any combination. The graphical objects represent logical structure, so this is an easy way to assemble large models quickly. The connections between the connecting pins on the objects can be made with a single line, with the actual connection resolving into tens of different connections to internal parts of the structure, and with information flowing in both directions along the conceptual link. This handles the bane of a dataflow assembly paradigm – when you are not sure which object should be first in the sequence (the circular reference of a spreadsheet). The connections between the objects ensure that the data flows are used to establish the appropriate sequencing. Sequencing through IF…THEN… statements in the model text can be used where the graphical approach becomes clumsy.

Particular Applications

Active Knowledge is applicable to many areas in the Financial Services industry. Some areas where the effectiveness of eCognition in handling the complexity has been demonstrated:

Asset/Liability Management

Portfolio Management

Risk Analysis

Dynamic Financial Analysis

Wealth Management

High Value CRM

Trend Analysis

Dynamic Financial Analysis

There have been many implementations of Dynamic Financial Analysis, but they have all suffered from the limitations of the system on which they were built – either spreadsheet or program. The underlying structure was neither flexible nor dynamic, no matter what the name implied, so it was not possible to analyse complex financial situations in any dynamic way. The system also had no ability to learn from its own operation, requiring every adjustment in strategy to be made with rules that could not take into account the complexity of the situations which occur in the simulation.


Computer-assisted (or fully computerized) business services must be capable of responding to considerable diversity. They must smoothly combine analytic and experiential knowledge, and rapidly alter their behavior in response to new information. The implementation of such systems requires an across-the-board improvement in a wide range of capabilities beginning with how knowledge is discovered, continuing through how it is used for simulation and analysis, and concluding with its deployment. Stated differently, the entire knowledge process must be integrated.

The implementation of point solutions to handle different aspects of the problem not only introduces integration challenges but may also sour management’s enthusiasm for the project by producing partial and unexciting initial results.

We propose that only eCognition offers the breadth of capabilities needed to build systems that overcome all current barriers and bring deals to their successful conclusion. No less importantly, Tupai’s quick implementations go a long way to getting projects up and running, demonstrating early successes and fortifying management’s support.

Unlike other approaches for capturing and deploying knowledge, Tupai’s technology enables transparency and verifiability of its workings, thereby supporting the controls and audits that are essential to responsible business management.

A presentation on Insurance Applications

Technical Discussion