Intelligensys automatic data mining, machine learning for product formulation

Developing better products faster

When you are developing new formulations, it may not be obvious which variables affect each property, and what rules govern your system. So, you might be measuring input variables that you don't need to or, even more worrying, missing ones you do. You might even be adding ingredients which have little or no impact on the observed properties, adding to the cost and complexity of your formulations.

If you don't understand the cause-and-effect relationships in your system, it's almost impossible to be confident about your data collection strategy. And without clear insights into the effects of making changes, it can be hard to define an efficient experimental program that meets your formulation goals.

In the past, formulators often used 3D graphs to try to understand what is happening in a formulation - with the real risk of missing an underlying trend. Neural computing avoids this, but traditional neural networks, which capture the cause-and-effect relationships accurately, are 'black boxes', so it can be difficult to understand their models. Now, FormRules automates much of the 'knowledge discovery' process in developing new formulations, presenting the information as easy-to-understand rules and graphs.

FormRules is 'data driven', so it is incredibly versatile. It can be applied to diverse formulation problems - in pharmaceuticals, health and personal care, home care, paints and coatings, adhesives... wherever formulation has an important role to play. And in each of those domains, it can be used for many different challenges. Provided you have your data, in a spreadsheet-type format, you can import it into FormRules.

And because FormRules can work over a network, it could facilitate information-sharing with your colleagues, improving the teamworking necessary to solve complex problems. It's an essential component of any knowledge management system for formulation.

To view a PowerPoint 'walk-through', follow this link. Or see some applications. Or just read on, to find out more about FormRules's technologies and features.

How FormRules works

FormRules tries out a number of models to see which one fits your data best, then refines that model to improve the accuracy. So, you can immediately see what formulation variables (ingredients or process conditions) affect each property. You'll know what to focus on, to drive the property in the direction you want.

FormRules uses neurofuzzy logic as its underpinning technology. This approach has been widely used in discovering new knowledge in financial and marketing applications, and is now being applied to formulation. Complex relationships, and subtle inter-relationships, are discovered within your data, without requiring you to pre-judge what form the relationships will take or indeed of having any knowledge that they do exist.

FormRules will work with even limited amounts of data. As with all modelling techniques, though, the better the data you provide, the more reliable the answer you will get. So, you can use FormRules to highlight what direction your experimentation should take. Once you've done more experiments, add them into the data set to develop new and better models.

Integrated statistics and graphics (both 2D and 3D) enable you to assess your models and communicate the results easily to your colleagues and customers.

For the most efficient product development, useFormRules in the early stages of your development, to see what is happening and to guide your experimentation. Once you have collected a reasonable amount of data, then use INForm to develop customer-specific optimal formulations.

Optional Support & Maintenance is provided with FormRules. We also offer training courses on its use.

To access the Application Notes showing how FormRules has been used, follow this link.