BETA
This is a BETA experience. You may opt-out by clicking here

More From Forbes

Edit Story

How AI Is Finding Patterns And Anomalies In Your Data

Following
This article is more than 3 years old.

One of the most widely adopted of the seven patterns of AI is the Patterns and Anomalies pattern. Machine learning is particularly good at digesting large amounts of data very quickly and identifying patterns or finding anomalies or outliers in that data. The “pattern-matching pattern” is one of those applications of AI that itself seems to repeat often, and for good reason as it has broad applicability.

The goal of the Patterns and Anomalies pattern of AI is to use machine learning and other cognitive approaches to learn patterns in the data and discover higher order connections between that data. The objective is to determine whether a given data point fits an existing pattern or if it is an outlier or anomaly, and as a result find what fits with existing data and what doesn’t. As one of the more widely used patterns, there are many ways in which this pattern is applied.

Digging deeper into your data

Data is at the heart of AI so it’s no surprise that computers excel at recognizing patterns in data. Whether it’s patterns of behavior, actions, input, or other patterns, AI systems are able to quickly spot it. Using artificial intelligence to spot patterns is ideal because humans, by nature, are unpredictable. AI is able to detect patterns that humans may not have even thought to look for. Also, artificial intelligence is able to pay attention to a lot more information at one time as opposed to the limited amount of data that humans can process and analyse.

Machine learning is all about using data and learning from it. Most of this learning comes from determining patterns inherent in the data. Rather than creating a program to tell a computer what to do with specific rules, machine learning allows a system to learn over time through examples and data. With programming, a human needs to set these rules. Therefore the system is limited by the number of possibilities programmed in. Machine learning on the other hand is not limited by such things.

There are many applications of AI in which you may want to use machines to spot patterns, or find anomalies and outliers in data. One widely implemented example of pattern or anomaly identification using AI is fraud detection. Fraud is simply defined as someone doing something they shouldn’t be doing. To find fraud an AI can look for actions that are not following the pattern of what they should be doing. If these actions look out of the ordinary the system can flag it for human review.

Another example that falls into this pattern is one that is used daily by many, but they may not even know they are using AI. When we use predictive typing on a computer or smartphone, this is powered by AI patterns. The computer looks at patterns in writing and is able to predict what word might be coming up next. Patterns of typing can become quite personalized over time to the point where the model is able to learn what specifically you are going to type next with a fair amount of accuracy.

Personnel and HR departments are also using AI to spot patterns in job applicants. The AI system is able to look at the applications and backgrounds of potential employees to determine potential good candidates and eliminate ones that don’t fit the job requirements. By using AI to help in the selection process, one would hope this would help screen candidates to move them to the next round as well as reduce bias in the hiring process. 

The patterns and anomalies pattern of AI can be seen in action in a variety of other ways. Intelligent monitoring, spotting mistakes or errors and making adjustments as needed, cybersecurity applications, and analyzing the stock market are all examples of some of the ways AI is being used to monitor patterns.

When letting the system find patterns on its own, it’s able to spot things humans might have otherwise missed. One example of this is what Walmart experienced in buying behavior around hurricanes. Walmart uses AI to detect sales patterns. One of the many trends they have detected is the connection between hurricanes and strawberry Pop-Tarts. It turns out that people go into Walmart just before hurricanes and in addition to stocking up on all the regular things such as water and batteries, they also stock up on strawberry Pop-Tarts. This insight has allowed Walmart to send extra truckloads of Pop-Tarts to stores in the hurricane's path. Unusual trends like this are sometimes hard for humans to spot, but something that computers excel at.

However, like anything that learns from data, you need to be careful about what the AI was trained on. Amazon came under scrutiny a few years ago after it was discovered their AI recruiting tool favored men for technical jobs. The patterns and anomalies pattern of AI, like the recognition and hyperpersonalization patterns are particularly susceptible to biased data sets. If you used bias data to train pattern recognition systems, it should come as little surprise that those systems will exhibit the same sort of bias as the training data. 

By thinking of AI projects in terms of the various patterns of AI, you can better approach, plan, and execute AI projects. Once you know that you’re doing a pattern and anomalies pattern, for example, you can gain insight into a wide range of solutions that have been applied to that problem, insights into the data that’s needed to power the pattern, use cases and examples of applications of the pattern, algorithm and model development tips, and other insights that can help speed up the delivery of high quality AI projects. These patterns help serve as a guide to help organizations do AI right and have a much greater chance of project success.

Follow me on TwitterCheck out my website