Most recent advances in artificial intelligence have been achieved by applying machine learning to very large data sets. Machine learning algorithms detect patters and learn how to make predictions and recommendations by processing data and experiences, rather than by receiving explicit programming instruction. The algorithms also adapt in response to new data and experience to improve efficacy over time.
Typically, machine learning is used to say something about a future event. Within the analytics domain you can distinguish between three subdomains: descriptive, predictive and prescriptive. The later two are the domain of machine learning.
Firstly, descriptive analytics is used to analyse something that already happened, for instance why were the sales of last month lower than expected?
On the other hand, predictive analytics is a typical feature of a data-driven organization. It will tell you more about what will happen in the future. Essentially it uses historical data to indicate something that has a high statistical probability of happening. E.g. what sales of product X do we expect this year based on the sales of the last five years?
Furthermore, prescriptive will not only tell you what will happen, but also what you need to do to make it happen. This method is used by leading companies in the tech industry such as Google and is for instance used in self driving cars.
A great example of how these last two domains are used is a recommendation engine. Platforms like Netflix use your behaviour and that of other users to recommend you a serie or movie. This model continuously learns by considering your new movie or serie choices. As a result, the recommendations might change over time if your interest shifts.
A machine learning algorithm uses training data and feedback from humans to learn the relationship of given inputs to a given output (e.g., how the inputs “square feet” and “number of bathrooms” predict housing prices)
How supervised machine learning works
1. A human labels the input data (e.g., in the case of predicting housing prices, labels the input data as “square feet,” “number of bathrooms” etc.) and defines the output variable (e.g., price)
2. The algorithm is trained on the data to find the connection between the input variables and the output. For instance, a large number of square feet and a large number of bathrooms indicates a higher price
3. Once training is complete, typically when the algorithm is sufficiently accurate, the algorithm is applied to new data to forecast the unknown price