Machine learning detects patterns in data, focuses on growth and adjusts when new data exposes a computer programme. It is closely related to computational statistics and links to mathematical optimisation. It offers tools that apply to predict, classify and cluster problems. It cannot be applied to everything and the data computer analyses depend on the problem entirely.
There are two popular approaches to machine learning :
- Artificial Neural Networks: It is a model of reasoning based on the human brain. Each neuron is an information processing unit and both data and processing are global and not local. Learning is through repeated adjustments.
- Genetic Algorithms: These are a class of random search algorithms based on biological evolution. These algorithms have an iteration known as a generation.
Four broad areas classify tasks solved by Machine learning depending on the nature of learning “signal” or “feedback”:
- Supervised Learning happens when labelled training data develop algorithms for known input and desired output, for example, spam filtering.
- Unsupervised Learning happens when data has no historic labels and the system has to find some structure within. The algorithms do not assume any desired output. It can also discover a hidden pattern in data such as item-based recommendation
- Semi-supervised Learning uses both labelled and unlabelled data. It is typically used for the same applications as supervised learning. Even though a small identified data uses a large unidentified data, for example, identify a person’s face on webcam.
- Reinforcement Learning involves an algorithm to decide actions with the greatest rewards through trial and error and find an ideal policy. To meet the greatest performance machines determine ideal behaviour automatically within a specific context, such as autonomous vehicles
Machine learning can also solve key information technology (IT) operations. It is used to analyse IT performance issues, offer insights and automatically find and isolate disruptions and failures.
To identify and predict anomalies machine learning depends on these types of data analysis:
- Descriptive (data mining): An analytic process to explore observations to discover meaningful patterns and/or quantifying data relationships between variables to find an approach. It solves questions such as what happened and why did it happen.
- Predictive (forecasting): To forecast some answer about unknown future events based on past behaviour. It integrates data and other inputs to predict what and when will an event happen.
- Prescriptive (optimization): Automatic process to find a better solution to a problem. It suggests choices for mitigating a risk or taking advantage of an opportunity.
