When were you born? Maybe the internet didn’t exist yet and nobody even thought of social media.
What is for sure is that Machine Learning still sounded as science fiction, although it was in its infancy already.
Machine Learning is a subfield of Artificial Intelligence research, based on the concept that machines do not only store and analyse data, but they can be trained to learn from it and to predict future outcomes by recognising patterns in the data.
Machine Learning has already entered all industries, from healthcare to finance. Small businesses and huge enterprises are investing heaps of money to develop state-of-the-art solutions which can set them apart from the competition.
Amazon has already launched Alexa, its voice-recognition system, whereas Facebook uses AI to detect mental illness.
HOW DO MACHINES LEARN FROM DATA?
All these technologies use models created by machines, based on their learning. In other words, Machine Learning uses different types of algorithms to analyse, categorize patterns in data and formulate predictions based on it.
More specifically, there are different types of learning algorithms, based on their learning patterns:
Supervised Learning: Supervised learning is a process of linking inputs to outputs: the computer is presented with an input and the desired output, and the goal is to find the function which best links the two pieces of data. Most of the times, it is a process of approximation rather than figuring out the true function.
With this process, Supervised Learning models relationships and dependencies between inputs and outputs, which can then be used to make predictions.
Unsupervised Learning: with Unsupervised Learning, machines are given an unlabelled set of data and are left to find hidden patterns in it. Contrary to supervised learning, outputs and inputs are not given: it is the algorithm’s task to find out patterns allowing researchers to arrange data on the two sides of the function.
Semi-supervised Learning: Semi-supervised learning falls between the previous two patterns, as only one portion of the data is labelled. With this algorithm, the machine is able to assign each unlabelled piece of data to a category (input/output) based on the labelled ones and to build models.
Currently, the most used method is Supervised Learning, which allows for a higher degree of certainty and is now applied in a wide variety of areas, from robotics to finance.
3 SUCCESSFUL MACHINE LEARNING-BASED MOBILE APPS
One of the fields where Machine Learning shows its utmost potential is Mobile Applications Development. The apps installed on our mobile phones collect an endless amout of information, which can be feed into the algorithms to build prediction models and improve user experience.
For example, ML is widely used in the development of Ecommerce Apps, which apply machine learning algorithms to provide customers with suggestions based on their tastes and past transactions.
Machine Learning algorithms select the best results in product search and provide recommendations on compatible products. Ecommerce apps are able to analyse purchase patterns, forecast trends and create specific promotions for customers so as to trigger purchase.
What is more, Machine Learning is also able to prevent fraud and unauthorised transactions by detecting abnormal behaviour patterns.
Another interesting example is the usage of Machine Learning in the development of mobile apps for smartphone customisation, based on their habits, preferences and usage patterns.
Indeed, Machine Learning is the basis of Prediction Launcher, an Android launcher allowing users to find the app they need when they need it. The algorithm analyses apps usage data and combines them with date and time of the day to provide the user with the app they are most likely to use at a certain time.
The more data is introduced into the system, the more accurate the prediction.
And what about the movies? I bet most of you spend endless nights watching Netflix, which was actually one of the first websites to exploit the potential of Machine Learning.
Thanks to a team of researchers tagging each and every piece of the movies in the Netflix catalogue, Machine Learning is able to combine tags with user behaviour data and to predict which show you will want to watch. Both the app and the web version use the same algorithm to formulate predictions based on your habits and preferences.
ONE LAST TIP
As accurate as predictions might be, they are not the only thing mobile users care about.
On the contrary, Amazon, Netflix, Tinder and all the other Machine Learning-based mobile apps would be uninstalled in less then 5 seconds if they lacked the key element that makes the difference between making it or breaking it: good user experience.
In fact, Machine Learning makes user experience all the more important, because it aims at creating customised experiences which cater for all the user's needs, while engaging them and increasing retention rate. The data collected can be used to predict what content the user will want to access and in what form, what combinations of contents they'll want to read/view and the path that will lead them there.
Each and every part of the UX design phase should be planned keeping in mind the transforming nature of Machine Learning and making sure your app is able to deliver adaptive content, i.e. "content that can, at each instance of use, change (adapt) – not just in appearance but in substance".
Netflix is excellent at offering adaptive content. Not only does it provide you with different suggestions every time you log in, but also the banners you'll see will change based on time and day. In this way, Netflix triggers customers' curiosity, while creating a customised user experience every time users enter the app.