The MindsI Engine is an adaptive machine learning system available within Fathom. The MindsI Engine has been designed to learn from the billions of records that you are collecting every day and to provide you answers.

The MindsI Engine requires just one step to setup and it will never get out of date as starts learning straightaway – that’s really important. No project is required, no batch learning, no weeks of time and resource and no paying vendors. It just starts doing its job as soon as soon as it is activated. The MindsI engine educes the need for expensive and time-consuming manual processes and the reliance on scarce analysts and experts.

Making Decisions – 24×7, 365 days a year

The MindsI Engine will listen to your data quietly in the background, 24×7, 365 days a year; helping to automate fraud decisions. This allows your experts to focus on just those decisions where human knowledge is needed.

MindsI can be combined with our Storm Rules Engine to make fully automated decisions perhaps sending a text message or email to a customer. Of course all this happens in real-time.

We think this is going to disrupt the way you think.

Get the right answers – fast

The uptake and use of neural networks and machine learning has been hampered by their “black box” nature; while they may give an answer we need to know why. All too often machine learning keeps the exact method of reaching a decision hidden and you are reliant on experts or consultants to set them up.

Using our patented technology (US 20110264612) knowledge is extracted from the MindsI Engine in the form of easy to understand English like rules. Understanding the underlying decision structure allows you to quickly determine Key Performance Indicators (KPI) in your business and because MindsI is always learning – you can see how these KPI’s are changing over time. We think that is amazing!

Out with the old

In a typical machine learning system, you will have to undertake a project:

  1. Choose and collect data
  2. Review and clean the data
  3. Convert the data into a suitable format
  4. Build models
  5. Evaluate models
  6. And finally deploy the chosen model.

This process takes both time and expertise; how quickly will the fixed model get out of date?

The MindsI Engine requires just one step and will never get out of date:

Setup and start

The MindsI Engine starts learning as soon as it is activated. The engine can process almost any type of input field:

  • Continuous values – e.g. value of a transaction, time at address, age
  • Discreet values – e.g. product codes, transaction type, country codes, region codes
  • Dates and Times
  • Unstructured data – e.g. emails, documents, voice-to-text

Just as the biological brain consists of a set of specialised structures for different types of inputs, such as vision, auditory, olfactory, so too does the MindsI Engine. Fields are automatically processed without fuss or time consuming human analysis.

Knowledge

MindsI needs to have sufficient experience of your data before it becomes sufficiently confident to start making decisions. The more examples it sees the more confident it will become and the more accurate. It is not just the volume of data but also how rich it is. MindsI constantly reviews its data feeds to ensure there is sufficient “intrinsic dimensionality” to enable an accurate output. MindsI has been created to only make decisions based on a solid foundation of learning – don’t ask MindsI to predict the outcome of a Formula 1 race based on just two examples or the price of Apple shares (AAPL) to within a fraction of cent in 90-days time (although we are working on this!).

We know that it is easy for models to make a decision – the trick is to maintain a high accuracy and reduce the number of incorrect decisions. These false-positives can clog automated fraud decision systems and can give them a bad reputation with their users and reviewers. A range of tools are provided to balance accuracy, false-positives (an “accept” when it should have been “refer”) and false-negatives (a “decline” when it should have been “accept”). Of course we do not expect you to work all this out – the MindsI Engine uses a novel Peareto Front Multi-Objective Evolutionary algorithm to sort it all out.