Big Data Analytics – Four reasons For Saffron’s Associative Memory Base For Predictive Analytics
Saffron introduces an associative memory system that is a combination of semantics and statistics, and it doesn’t require authoring for ontologies, or semantics, or employing PhDs to model the statistics.
Consider the typical B2C scenario. Let’s say you’re the CMO of a company, and you want to understand what’s going on the internet, and you want to combine it with your financial data, you won’t find many Big Data Analytics tools on the market that one can use out of the box for predictive analytics of hybrid data (structured enterprise data and unstructured web data). This last point is critical in the modern world of big data where one has to combine data from the internet with corporate data.Here are four reasons Saffron’s Big Data Analytics beats traditional BI approaches:
1. Statistics can’t handle diverse data well
Take the challenge of big data and National Security. One can understand a sleeper becomes a terrorist using Saffron but if one tries to look at it statistically, one will fail because once one has statistically significant or relevant data, it’s already too late.
The same is true for customers. Potential customers. Returning customers. Churning customers. Statistics give the answers too late. One needs to know before the customer churns – not after.
2. Predictive Analytics needs to handle transactions
It’s easy to see why one would want to be transactional taking fraud detection for example.
Let’s think about insurance claims or the IRS.
Assume you have trained your system; you understand what fraud means. You get in an insurance claim or you get a tax return, and then you want to look at the return and you want to see if it’s valid or potentially fraudulent …
You can try to compare existing patterns, but a new claim may form a new pattern, and since it is the first case to attempt to cheat in this new way you don’t know it yet.
If you use a batch technology like MapReduce (Apache Hadoop) you would have to run the whole process for discerning the fraud patterns again; for example after 100 new claims have arrived.
And here you see again that statistics wouldn’t work. One would need many new claims (at least some 100) to discern new fraud.
So this is the nice thing about Saffron’s Associative Memory system – it is transactional and finds outliers and emerging patterns out of the box.
3. Saffron’s Big Data Analytics support front line decision makers
One strength of associative memories is, they know what they don’t know.
They can handle very sparse and diverse data. Looking at existing connections they can establish relevance, and they can see if there is a new fraud pattern emerging in a new insurance claim or a new IRS tax return.
Saffron’s Associative Memory Base can recognize that a return is outside of the established patterns.
If there’s a question as to whether it should be flagged, Saffron’s Associative Memory Base can also flag it for a human to decide.
4. The knowledge of time – anticipating the future and forgetting the past
Saffron’s Associative Memory Base also understands time; that’s very important.
Google’s traditional statistics approach (inverted index) for example doesn’t distinguish before and after.
But Saffron does… It has time built in. One can use it to anticipate future events and distinguish them from past events.
Associative Memories change over time. They are schema-less and thus can forget and learn new connections on the fly.
That’s why it’s called memory.
Saffron forgets, like we do with our biological memory. In our brains, we forget things that we haven’t touched a lot of times, and other things become more up-to-date, more important.
These are just four reasons why Saffron delivers customers a significant advantage for Big Data Analytics.