We touched on it in a previous Target Tech Bytes, but it’s a topic that warrants its very own edition. This month we focus our lens on machine learning and take a closer look at this impressive technology and how it’s opening new realms of possibility.
In an industry of abundant data, where the demand for digital self-service has erupted and financial services firms have no option but to take digital transformation seriously, machine learning has emerged as a powerful tool for analysis, prediction, and decision-making.
So, what is machine learning?
Machine learning is a subfield of Artificial Intelligence (AI) that focuses on creating algorithms and models with the ability to learn, identify patterns and make predictions or decisions with minimum human interaction. All of this is achieved without explicit programming. Machine learning uses algorithms, essentially its building blocks to learn through experience and example and improve over time.
A key feature of machine learning is its dependence on data. The more data available for analysis, the more accurate and robust the machine learning models can become.
When was machine learning invented?
The concept of machine learning was conceived in 1943 when a paper published by logician Walter Pitts and neuroscientist Warren McCulloch attempted the mathematical modelling of neural networks.
In 1950, Alan Turing asked the question, ‘Can computers think?’ in a paper called Computing Machinery and Intelligence. He set the groundwork for modern computing and introduced the Turing test, which determined a machine’s ability to demonstrate human behaviour.
These events paved the way for further milestones and the development to what machine learning is today.
The four approaches to machine learning
To understand how machine learning works, you first need to understand the methods and algorithms, essentially a set of rules, that machines use to make decisions.
The most common types of machine learning are:
- Supervised learning – The most popular approach to machine learning, supervised learning algorithms learn by example. They rely on sample data, also known as training data, which consist of inputs labelled with the correct outputs, to search for patterns and learn. Predictions are then made on new data based on the training data the model has learned from
- Unsupervised learning – These algorithms analyse unlabelled data to discover patterns and important features. They’re not trained with the correct answer, so need to detect patterns to make decisions based on their findings and predictions
- Semi-supervised learning –The algorithm analyses both labelled and unlabelled data. The labelled data is used as an input to learn and make predictions for the unlabelled data
- Reinforcement learning – These models are designed to establish the best course of action based on the scenario given. As no training data is provided, they do this through trial and error, learning from their mistakes to make the next decision, the one that’ll give the most reward
- Deep learning – A method of machine learning that uses layers of algorithms that create artificial neural networks. These can learn and make intelligent decisions without human intervention. The artificial neural networks are inspired by the structure and function of the human brain’s neural networks and are designed to learn from huge amounts of data.
Machine learning made simple
An example of a machine learning application that most of us can relate to is email spam filters. Based on past analysis, an algorithm is used to review incoming emails and determine whether they’re to be moved to spam folders.
Product recommendations on websites and friend recommendations on social media are also examples of machine learning. Through enabling cookies these sites can analyse and track the way you use their sites to make predictions and recommendations personal to you.
Pros and cons to using machine learning
Machine learning is not without its challenges and limitations, this can also differ based on how it’s applied. Let’s look at some general pros and cons of this technology.
- Machine learning can be used in many different industries, including education, healthcare, and finance, automating complex tasks and processes that would otherwise require significant human time and effort. A research study found that an AI model outperformed doctors when reading mammograms in the diagnosis of breast cancer
- Machine learning can analyse massive amounts and different varieties of data, helping businesses make data-driven decisions from valuable predictions
- Machine learning algorithms learn from historical data, and their performance improves over time. They can identify anomalies or patterns that can often be missed by humans, which is particularly valuable in a department such as fraud detection
- The room for improvement in machine learning is enormous, the field is evolving at pace with new innovations occurring regularly, making it one of the leading emerging technologies.
- Machine learning is still open to error, for example, if the sample data provided was mislabelled or biased, this will lead to incorrect and/or biased models
- Implementation and management costs for machine learning software can be extremely high and often require significant computational resources. This can be a barrier to most businesses
- Poorly managed machine-learning models could expose firms to reputational, regulatory, and financial risks.
Machine learning in financial services
In 2022, The Bank of England and The Financial Conduct Authority (FCA) rolled out a joint survey to understand the use of machine learning within the financial services sector. Out of those who responded, 72% of firms reported using or developing machine learning applications. Yet the biggest constraint to successfully implementing the technology is legacy systems.
Fintech firms and challenger banks are making strides in the digital automation space by using machine learning to streamline processes such as credit score assessments and mortgage applications, making them more efficient and personable to the customer.
The likes of Monzo are using machine learning to assist with their fraud detection process, significantly reducing the amount of fraudulent top ups on their prepaid debit cards.
Peer-to-peer lending company, Zopa have created a tool called Borrower Power which uses a combination of machine learning and artificial intelligence. It provides their customers with bespoke actions to improve their credit score and loan products best suited to their needs based on data analysed by the model.
How can machine learning benefit your business?
Data is the beating heart of most firms. Without it they'd struggle to operate effectively. Machine learning can analyse great swaths of data to generate valuable business insights. These can enable better decision-making and translate into actionable strategies that support business growth and enhance customer experiences.
By leveraging the power of machine learning, while maintaining human ethics, your business can unlock its true potential where humans and machines can work together in harmony.