You would be hard-pressed to find anyone remotely connected to the tech world who hasn’t heard the term “machine learning” recently. Technological innovators are constantly praising machine learning techniques and countless articles (including ours) deconstruct the newest applications based on machine learning strategies. But what is machine learning, really?


In very simple terms, machine learning is “a type of artificial intelligence (AI) that provides computers with the ability to learn without being explicitly programmed. Machine learning focuses on the development of computer programs that can change when exposed to new data.” (TechTarget) For a more visual representation of how a machine learning algorithm works, here’s an example of one learning how to beat a level in Super Mario:

Machine learning algorithms can get very complex and are used for various pattern recognition processes like classification, clustering, density estimation and dimensionality reduction problems, to name a few.

Many of these machine learning applications are well-known in the world of statistics, so why is it getting so much hype recently? Generally speaking, there are two reasons for the recent swell of machine learning love:

  • Firstly, the quality of its results correlates directly with the amount of data computers can store and process. Since computers are rapidly and consistently increasing the amount of data that can be stored and processed, it is just recently getting to the point where it can yield impressive results.
  • Secondly, the sheer amount of data available in today’s digital world makes it possible for machine learning to yield actionable insights.


Machine learning algorithms are divided into three groups:

1. Supervised learning
  • Where input and the output are available and the algorithms are trained to fit the input to the output.
  • Example: Handwriting recognition, where the output (the written alphabet) is known and the input (handwriting) must fit it.
2. Unsupervised Learning
  • Where only the input is available and the goal of the algorithms is to find the structure in the data on its own.
  • Example: Network security, where the algorithm is tasked with profiling presently unknown network threats and blocking them in the future.
3. Reinforcement learning
  • Where in a dynamic environment the algorithms learn to perform a goal, by receiving positive and negative feedback.
  • The Super Mario example above is a perfect example of reinforcement learning, the goal being the highest “fitness” score possible.

In all three cases, the algorithms learn from their experiences. When new data is introduced, results for the data is predicted from the previous learning processes and further learning is achieved through the new data. With this iterative process, models are able to independently adapt. Independent problem-solving skills and learning from data is generally described as artificial intelligence.


Applications of machine learning vary widely and are found in several devices and applications we use today. Examples included recommendation engines, spam filters and speech recognition programs. In the near future, it will even be one of the key processes in self-driving cars.

In Financial Services, banks and businesses currently just use machine learning for two main purposes:

  1. Insights from data
    1. Its algorithms indicate the best investment opportunities and derive the best moments to trade.
  2. Fraud prevention
    1. In general, it’s used to find patterns in data, but it’s also able to find anomalies. It is therefore perfectly equipped to find and prevent fraud.

In B2C Marketing, machine learning is an often-used discipline. Recommendations of items you might like while shopping online are powered by machine learning techniques. The algorithms personalise each costumer’s shopping by analysing patterns, either from their own buying history or from all customer behaviours. Predictive Marketing is a significant improvement to sales and marketing strategies.

Machine learning is a lot less commonly used in B2B and is just recently being adopted by tech-savvy businesses. Where it primarily comes in handy in these instances is through the detection of patterns in user actions across multiple channels of input. For instance, machine learning AI can detect when a prospect clicks on an email regarding a certain fund as well as when they click on similar funds on your website. From that input it learns and infers that person has an interest in those types of funds, and automatically sends them updates on similar funds in the future.

Although this tech is still gaining its footing, its a safe bet that it, and tech like it, will be the next step in learning and predicting individual customer behaviour on a mass scale and will completely revolutionise how you market your funds.


All these examples demonstrate the incredible potential machine learning has to offer. The development of machine learning algorithms makes it possible to quickly and dynamically derive models that detect patterns in extensive sets of data. Organisations can use machine learning to avoid risks or to identify opportunities for profit. Plus, it creates Google Artificial Intelligence Dream images. Either way, it’s pretty awesome.