Machine learning: value for companies

Machine learning (ML) algorithms allow computers to define and apply rules that were not explicitly described by the developer.

There are many articles dedicated to machine learning algorithms. Here is an attempt to make a “helicopter view” description of how these algorithms are applied in different business areas. Of course, this list is not an exhaustive list.

The first point is that ML algorithms can help people by helping them find patterns or dependencies that are not visible to a human.

Numerical forecasting seems to be the best known area here. For a long time, computers were actively used to predict the behavior of financial markets. Most of the models were developed before the 1980s, when financial markets gained access to sufficient computing power. Later these technologies were extended to other industries. Since computing power is cheap now, even small businesses can use it for all kinds of forecasts, such as traffic (people, cars, users), sales forecasts, and more.

Anomaly detection algorithms help people scan a large amount of data and identify which cases need to be checked for anomalies. In finance they can identify fraudulent transactions. In infrastructure monitoring, they allow you to identify problems before they affect the business. It is used in manufacturing quality control.

The main idea here is that you should not describe every type of anomaly. You provide a large list of different known cases (a learning set) to the system and the system uses it to identify anomalies.

Object grouping algorithms allow you to group a large amount of data using a wide range of meaningful criteria. A man cannot efficiently operate with more than a few hundred objects with many parameters. The machine can make grouping more efficiently, for example, for customer / lead qualification, product list segmentation, customer support case classification, etc.

The recommendations / preferences / behavior prediction algorithms give us the opportunity to be more efficient when interacting with customers or users by offering them exactly what they need, even if they have not thought about it before. The recommender systems work really bad in most services now, but this sector will improve rapidly very soon.

The second point is that machine learning algorithms can replace people. The system analyzes people’s actions, builds rules based on this information (that is, it learns from people), and applies these rules by acting instead of people.

First of all, it involves all kinds of standard decision-making. There are many activities that require standard actions in standard situations. People make some “standard decisions” and escalate cases that are not standard. There are no reasons why machines can’t do that: document processing, cold calling, accounting, front-line customer support, etc.

And again, the main feature here is that ML does not require an explicit definition of rules. It “learns” from the cases that people already solve during their work and makes the learning process cheaper. Such systems will save business owners a lot of money, but many people will lose their jobs.

Another fruitful area is all kinds of data collection / web scraping. Google knows a lot. But when you need to get aggregate structured information from the web, you still need to entice a human to do so (and there’s a good chance the result isn’t really good). The aggregation, structuring and cross-validation of information, based on your preferences and requirements, will be automated thanks to ML. People will continue to perform qualitative analysis of the information.

Finally, all of these approaches can be used in almost any industry. We must take this into account when predicting the future of some markets and of our society in general.