Predicting The Future with Predictive Models

Interview conducted by G.P. ThomasAug 22 2012

In this interview, Dr Don Perugini, Director of Intelligent Software Development (ISD), talks to AZoCleanTech about the environmental need for predictive models and how accurate these can hope to be.

Could you please provide a brief introduction to the industry that Intelligent Software Development (ISD) works within and outline the key drivers?

ISD works in three key sectors: analytics, clean technology and cloud computing. Our software is about predicting the behaviour of people and populations, and therefore can be applied globally to consumer problems, such as water, energy, transport, waste, retail, health, and finance. To date, we have worked with businesses and Govt in Australia, Europe and USA in the water and energy sectors, assisting with strategy and policy.

Key drivers for our clients using our technology is their limited ability to integrate complex data from their existing analytics technologies to better represent the complexities of human behaviour in order to facilitate accurate forecasting and decision support. Commonly used analytics technologies such as data mining and statistics are focussed at observing what people did in the past, or providing average/simple predictions. But the greatest business value can be realised from emerging technologies such as simulation, which can provide better predictions, provide better assessments of how to respond to a range of complex business situations, and provide insight into the consequences of those decisions.

Could you please give a brief overview of ISD, including the history and the development of the company?

ISD has cloud based simulation consumer analytics software, which we call SimulAIt Online or SOL. SOL can simulate a complete population of mass-consumers to a high level of detail so that business or Govt can more accurately predict how people will behave, but more importantly test options (through what-if scenarios) for influencing their behaviour. As you can imagine, access to a smarter analytics tool that can provide better insight into what strategies and policies can influence a diverse range of consumers' buying decisions or behaviour can be very valuable for organisations.

ISD commenced commercial operations late 2007, with the founders Dr Don Perugini and Dr Michelle Perugini originating from the defence and health sectors, respectively. ISD has grown to 7 staff, mix of full and part time, and anticipates further growth as ISD expands globally in a diverse range of markets.

How is ISD unique within the field of predictive simulation?

Most analytics technologies that are being used are called descriptive or predictive analytics, such as data mining and statistics, which process large amounts of raw data to provide insight into what has happened in the past, or provide simple forecasts or predictions.

Emerging analytical tools, and what Gartner describes as having greater value for businesses, is simulation which can utilise the complex data from existing analytics technologies to create a rich simulation environment that allows decision makers to not only predict behaviour, but allow testing of strategies to influence behaviour, and provide insight into the consequences of those decisions.

SOL also overcomes limitations with traditional analytical tools to integrate complex data to better represent the complexities of human behaviour. Unlike other approaches, SOL can validate the accuracy of its models (achieving over 95% accuracy). Since SOL uses no or very little demand data to configure models, the model can be "run forward from the past" to show that it can accurately forecast (past) demand data (i.e. predict client’s CRM data) without using this data. This provides clients confidence that the model works and can accurately predict consumer behaviour.

Lastly, a range of users can use the software using the collaborative web-based application.

Could you provide examples of environmental situations in which your forecasting can be utilised?

To date, we have worked with businesses and Govt in Australia, Europe and USA in the water and energy sectors, assisting with strategy and policy, designing smarter cities, etc., for example:

• assisting our clients understand change in behaviour and demand for situations they haven't experienced in the past (and dont have historical data for);
• assist with pricing review and capital planning;
• assessment and design of demand management and behaviour change programs, including incentive programs (e.g. retro fit) and rebates to help consumers reduce their water and energy footprint;
• (micro)-marketing of programs and customer personalisation;
• and energy load forecasting, using data from smart meters/grids to assist with hedging and risk management.

We have achieved very good results with our projects, with accuracy over 95%.

Your predictive solutions are underpinned by the micro-simulation platform SimulAIt-could you describe how this works and how it utilises Agent Based Modeling?

SimulAIt uses cognitive software from the Defence sector, integrated with Census data, to better represent human decision making to enable Government and business to accurately predict consumer response to new policies, products, prices, marketing, and competitive strategies.

Cognitive agents are software components within our simulation environment. Agents in our simulation environment represent consumers and their prescribed rules to mimic the complex decisions, behaviours, and social interactions of different consumers. We extrapolate these behaviours and logic across Census data, and generate many agents representing the many different/diverse types of consumers, which can results in the accurate simulation of millions of consumers in order to predict their emergent behaviour.

In any real-world situation there are an almost infinite number of possible scenarios and parameters-how does your software ensure accuracy and how many assumptions are made?

As discussed above, unlike other approaches, SOL can validate the accuracy of its models (achieving over 95% accuracy). Since SOL uses no or very little demand data to configure models, the model can be "run forward from the past" to show that it can accurately forecast (past) demand data (i.e. predict client’s CRM data) without using this data. This provides clients confidence that the model works and can accurately predict consumer behaviour.

Using ISD’s SOL web-based platform, our clients can view and change all the parameters, trends and assumptions used to drive the logic in their simulation scenario/forecasts. In this way, they can run what-if scenarios to test a range of business problems and strategies and assess the likely future outcomes.

Could you tell us about an organizational case study that you were particularly proud to be involved with?

Various case studies can be found on our website.

ISD has recently been involved in the Australian Cleantech Competition 2012-do you feel enough is being done to promote growing cleantech companies?

The CleanTech competition has greatly assisted the promotion of the cleantech sector, and thus we are delighted to participate in this competition and allowing us to leverage off the exposure and promotion. This has been very valuable for us.

In your opinion, what is the current attitude towards clean technology in Australia? Is this different to other parts of the world?

We have been involved in a project with Veolia in Europe, and had the privilege to be invited to attend a smart cities conference with them. What we found is that in Europe large global companies such as Veolia promote the fact that they work with new innovative companies in the cleantech sector to address some of our most challenging environmental problems. This is the type of culture we would like to see more of in Australia.

How do you see forecasting software progressing in the future? Will it become more accurate? Can it potentially be used to alter people’s behaviour to become more sustainable?

With advances in data capture, analytics and cloud computing, we see that more sophisticated and emerging analytics such as simulation can be more commercially applied to business of all sizes in a large range of industries. As these technologies develop further, and more data is collected, such as with Smart Grids and Smart Water, these technologies are likely to become more accurate and functional (i.e. address a range of business problems). This will better inform Govt and business about how they can influence people’s behaviour to achieve better environmental outcomes.