patterns and creating predictive models or multi-objective optimisation projects, and even execute virtual experiments using predictive models. In addition, the DataStories automated machine learning approach aims to equip these domain experts with the ability to easily operationalise their findings and share the results with their peers and management. Using this technology allows lab users

to gain insights into which experiments are worth doing the most to increase the body of knowledge, and which ingredients and conditions have the most positive effect on the desired outcome. Embracing this approach can help to accelerate time to discovery, shorten development cycles and improves success rates. One example can be found in a workflow that was highlighted by Smits. Shown in Figure 2, this workflow ‘has been very effective in running projects from product development such as catalyst design, coatings, detergents, polyurethane and epoxy formulations,’ notes Smits. ‘In a typical project, one will loop through

this workflow several times, applying augmented analytics tools at every stage, reusing and augmenting all available data at all times to optimise the balance of what can be predicted from existing data using modelling and, for example, virtual design of experiments versus which and how many additional experiments need to be carried out to complete the models for optimal performance,’ said Smits. One of the key themes running throughout

 Figure 2: augmented analytics workflow

the design and implementation of this augmented analytics platform is the idea of democratising the use of analytics – making it easier for domain scientists to generate insight from the data they create. This democratisation is a key choice that has been at the heart of the development of the platform from DataStories. ‘This software allows a researcher to upload a table with possibly thousands of variables, specify one or more KPIs – variables that are the targets to be optimised – and press go to generate a complete data-story that gives you the correlation structure and dependencies between all your variables. Also, a robust predictive model is automatically built with the minimum set of necessary and sufficient variables to give you maximal

predictive power. It shows you the what-if explorer that gives you quantifiable insight in the behaviour of the model, and shows you potential outliers in the data that do not satisfy your model,’ stated Smits. ‘At no point does the user have to answer

any questions on which algorithm to use, what parameter settings are optimal, what to do about missing or badly scaled data, or how to deal with outliers. ‘The focus is on turning the data into actionable information quickly, in a form that can be easily digested and shared with colleagues and the rest of the organisation. The R&D person is expected to apply his/ her domain expertise to decide on what action to take next, but it doesn’t need a data scientist to do this,’ he concludes.

New Whitepaper: Augmented Analytics in R&D Available online now

We live in a time of unprecedented innovation. Product design cycles keep getting shorter, personalized and custom products are becoming the norm and external requirements are changing constantly. On top of this, global problems like climate change, clean energy, water, phosphorus and nitrogen flows, plastics waste and other planetary boundaries that are being transgressed are calling for redesign of almost everything we are familiar with today.



*Registration required


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32