How machine learning could change my business — and yours
If you want to stay right on the cutting edge of new technology, try teaching. I learn constantly from my MBA students at Chicago Booth. This quarter, I’m mentoring a team as it explores the impact the cloud will have on business and society. It’s not a subject I likely would have studied without them, but teaching this course gives me invaluable insight into a technology that’s transforming the business world.
The team is spending much of its time investigating how the cloud can support machine learning, an important aspect of artificial intelligence. Our expert has taught us that machine learning is useful for four main activities: regression, classification, dimensional reduction, and anomaly detection.
Regression refers to a way to model the relationship between different variables, allowing you to make better and better predictions. Machines’ ability to analyze large amounts of data allows them to continually refine the regression, bringing us closer to understanding the correlation between any set of variables. Real estate offers a great example. Instead of guesstimating what the list price for your house should be, a regression analysis can tell you at exactly what price your house is likely to sell by identifying and weighing the most important features of your home, then tying those to market data.
Classification is about pattern recognition — categorizing a new observation by using a set of data with known categories. This is what helps doctors identify tumors. By analyzing a tumor scan against the data from a vast array of tumor diagnoses, machines can categorize a tumor as benign or malignant with a fair amount of accuracy. The more data the machine has to draw from, the more accurate its classifications can be.
Dimensional reductions is another way of saying clustering — discovering the inherent structure in data so you can summarize it using less information. The example here comes from marketing. Anyone seeking to deliver a message to a targeted audience derives great benefit from identifying the key factors that differentiate groups of customers. Dimensional reductions allows them to bucket customers efficiently by focusing in on the data points that are most likely to determine whether or not someone will be interested in the product or service being advertised.
Anomaly detection allows us to find data points that are unusual. Machines can learn what normal activity looks like and rapidly identify anything that is notably different. This is especially relevant to IT, where machines can become very skilled at identifying security threats and fraud.
Any or all of these four processes might be a part of your business already. In management psychology, I’m regularly applying these means of analysis to behavior, to help my clients better understand why they’re doing what they’re doing and what actions they need to avoid or encourage in order to improve. As machines make these activities easier to perform and access to data makes them more accurate, I can already see how I can use these four means to surface insights at an organizational or industry level. I’m sure many of you can think of applications that would push your own businesses forward.
There are other consequences to cloud-driven machine learning. For one thing, it means that many jobs that focus on data analysis are going away. As machines are exposed to more and more data, their power to understand and manipulate that data is growing rapidly. As with any powerful tool, the implications are both exciting and scary. But whether you’re a techie or not, understanding what is happening in this arena is critical for remaining relevant.
