According to a Netflix study, up to 75% of people watch what is suggested to them in the suggestion reel. These suggestion reels use machine learning algorithms to determine common viewer preferences and behaviours. Many applications used by people everyday harness machine learning algorithms. Gmail, for example, uses it to filter spam and credit card companies utilise it for fraud detection.


So what is machine learning?

The simplest way to define machine learning is that instead of instructing a computer what to do, it’s able to determine by itself what it needs to do with the data provided. It’s a move away from static programmed computation, where a computer is programmed to do one action only, to one that is driven by data. Machine learning can help computers to uncover hidden insights without having to be explicitly programmed on where or what to look for.

The science of machine learning is not new. For example, Amazon has been using machine learning to provide hyper-personalised product recommendations for years. However, with the ability to store big data cheaply and calculate faster, machine learning is gaining momentum.

Machine learning algorithms

One area of machine learning is reinforcement learning. It includes areas of deep learning and neural networks used for driverless cars and playing the complex strategy board game, Go. However, there are other machine learning algorithms that require more human direction.

Machine learning algorithms outside of reinforcement learning include supervised and unsupervised learning. Both can apply to different business applications and use their own algorithms.

Supervised learning is a function where outcomes are inferred from the example input data and historical outcomes. It also analyses and solves regression problems that determine which factors influence an outcome. An example includes finding what characteristics of a house such as size, number of bathrooms, etc. most affects its price (linear regression) or what information recorded as part of a sales opportunity, affects its chance of being won or lost. There are a number of algorithms that are used for supervised learning and each is best used in specific instances.

Unsupervised learning looks for hidden structures from unlabelled data (clustering). This involves feeding a model unlabelled data and asking it to perform pattern recognition. An example of this is clustering the distinct groups your customers fall into based on buying behaviour.

Historically, the computation power required to undertake large scale machine learning to process big data has been prohibitive to many companies without such resources. Today, companies with the required resources such as Amazon, Microsoft and Google are offering these services and platforms.

Some offer “out-of-the-box” services based on algorithms trained on large data sets such as image analysis to identify inappropriate photos, speech recognition to read text to a person with vision impairment and sentiment analysis in text. Google uses sentiment analysis to ensure holiday ads for example, are not displayed alongside articles with a negative sentiment. Some also offer platforms that allow you to create your own machine learning algorithms and provide hosting such as Azure Machine Learning studio.

How can you benefit from machine learning?

Have you been collecting big data on your customers but don’t know how to make sense of it? Machine learning will enable you to determine patterns in your customers’ buying behaviour.

Are you looking to have some custom software developed? Augment your custom software with data analytic components to identify events such as near misses or user habits in customer facing applications.

Do you currently use a business rules engine such as InRule? Machine learning and rule technology can be used together to help you make decisions on outcomes faster, with greater accuracy than was ever possible in the past.

The possibilities of machine learning are endless. It is the future of software development. Machine learning can produce high value predictions that can help you make better business decisions, in turn leading to reduced costs and increased revenue.