Science to solutions
Presage uses cutting-edge science, both proprietary and public, created by its scientific team to turn science into solutions for your organisation. With over 150 publications in Complex Adaptive Systems, Evolutionary Computation, Machine Learning, Spatial Data Mining and Complex Inference Networks, as well as their applications to financial markets, obesity and chronic-degenerative diseases, COVID-19 and other emerging diseases, biodiversity, crime, disaster risk, homelessness and politics, among others, as well as a wealth of successful business projects, Presage’s data science team gives our clients access to both the creativity of academic research and the proven ability to solve of real-world business problems.
Complex Adaptive Systems
Your business is both complex and adaptive. Complex in that what happens in it, and to it, depends on many, many factors – the problem of “multi-factoriality/multi-causality”. Adaptive in that what works today may not work tomorrow – the world changes. Dealing successfully with these characteristics is on the very forefront of scientific research. The Presage team has more than 20 years experience researching Complex Adaptive Systems, developing new methodologies and testing them out on real-world problems. The results have been incorporated into Presage’s HIPRE Platform, which has the capacity to create Bayesian prediction models with almost unlimited number of factors – complexity – as well as allowing the models to learn in the presence of new data – adaptation.
Asset allocation, production line optimisation, route planning, scheduling and many others, are all problems where there are an exponential number of possible solutions. Usually, Human Intelligence is used to search the space of possibilities using intuition and experience as a guide. However, besides being biased, Human Intelligence can only explore a tiny fraction of the possible solutions, leaving the possibility that much better solutions lie undiscovered. Evolutionary Algorithms, based on the principles of Darwinian evolution, are an area of Artificial Intelligence that permit a principled exploration of the possibilities that is far beyond the capacity of Human Intelligence. The Presage team have been a leader in the development in this area with several proprietary algorithms contained in its HIPRE Platform.
Machine learning is a more suitable label for many of the most exciting developments in Artificial Intelligence in recent decades, capturing the idea that, as with a physical “machine”, a machine learning algorithm is designed to do one thing, but only one thing, very well. The thing that machine learning-based prediction models, P(C|X), are designed to do well is to predict the outcome C given the presence of the factors X. There are a very large number of machine learning methodologies, ranging from simple rules-based systems to the most sophisticated deep learning methods. The Presage team has ample experience in many of these methodologies, though the HIPRE Platform is based on Bayesian learning algorithms which we believe are the most suitable for dealing with Complex Adaptive Systems.
Some of the most exciting recent developments in Artificial Intelligence in recent years have been in the area of Deep Learning, and most recently, in the development of Large Language Models. These models can be characterised as “black box” in that the relation between output and inputs – the WHYS of your prediction – is very opaque. For many of their applications, such as translation, transcription, voice and image recognitions, the WHYS are not important. However, when considering an organisation’s Decision Cycles, the WHYS behind the prediction are vitally important. Presage uses Deep Learning, when appropriate, as a component in its Decision Cycle modelling, combining it with more white box prediction models to combine the best aspects of multiple machine learning algorithms.
Spatial Data Mining
One of the main areas of data generation in the Data Revolution is spatial data associated with the business question – WHERE? Many GIS (Geographic Information Systems) are available for a descriptive analysis of such data but without the capacity to produce prediction models. Presage has developed multiple innovations in this area that allow for the prediction P(C|X) of any class of interest, C, represented as a geographic distribution, as a function of any number of predictors X that are also represented as geographic distributions. These developments have been successfully applied to the modelling of the risk factors for homelessness and unfair housing, for automobile accidents, for emerging diseases such as COVID-19 and for crime prediction and mitigation.