What is Machine Learning?

Machine learning, a type of AI, uses a data analysis method that automates model building by gathering and interpreting large sets of data.

In this modern, increasingly competitive business market, reliable insights into customers and emergent trends can mean the difference between success and failure. To meet this need, businesses turn to data analytics. Machine learning (ML) applies advanced AI solutions, using data and algorithms to create data models. A model is a mathematical expression that approximates the relationship between the variables that appear in the data and enables the ability to predict one from the other. A very simple example of a model would be a linear relationship that predicts a person's weight from their gender and their height. Rather than following specific, pre-programmed rules, machine learning mimics the human learning process, improving itself through experience and training.

Using machine learning solutions to build precise models empowers organizations to analyze extremely large, complex data sets, delivering faster, more-accurate results at scale. With machine learning, businesses enjoy detailed insight into opportunities, risks, and customer needs. And while this often means improved returns, the true possibilities of machine learning may be almost limitless.

Artificial intelligence (AI) 

AI is a discipline that is used throughout information technologies to describe any attempt to replicate human or near-human intelligence in machines. AI encompasses both machine learning and deep learning.

Machine learning  

The term "Machine Learning" is typically used to refer to classic data-based algorithms that identify patterns and perform tasks like classification, regression, and clustering— The more information it has, the stronger it will perform.

A model is specified by several parameters. The concept of training an ML model means it moves to optimize the parameters while trying to minimize error between predictions and the actual truth values that appear in the data.

Deep Learning 

Deep Learning is a younger field of AI that is based on neural networks. It is a subset of Machine Learning, using and structuring parameters in connected layers to create artificial approximations of human neural networks.

Training a neural network requires large amounts of data and computational resources, but the resulting models are often much more powerful than those obtained with classic machine learning algorithms.

Machine learning results in computer algorithms to transform data into intelligent interpretations and actions. Data mining hunts for actionable intelligence in existing, available data.

Data mining falls more under the umbrella of business analytics. It focuses on teaching computers how to identify unknown patterns, anomalies, or relationships in a large data set. Humans can then solve problems using this data. The process is more manual and usually requires human intervention for decision making.

Machine learning falls under the AI umbrella and focuses more on teaching a computer how to learn analysis of large datasets and its patterns. After initial programming, machine learning is capable of learning and improving without the need for human intervention. The computer becomes more intelligent and grows by itself, in a way, rather than being reactive and simply analyzing data it is given.

Machine learning generally follows a specific process, outlined below:

  • Gathering data
    Reliable data is collected so that it can then be used to inform the predictive model.
  • Preparing data
    Collected data is pulled together, irrelevant details are removed, and any necessary adjustments are made (such as correcting errors, removing duplicate data, etc.). Data is split into two sets: training data, which is most of the dataset and will be used with the machine learning model, and evaluation data that is used to test the effectiveness of the model after it has trained.
  • Choosing a model
    A model is selected. Many different machine learning models exist, and some are better suited to specific use cases than others.
  • Training
    The refined data is used within the chosen model to incrementally improve that model's predictive ability.
  • Evaluating
    After the model has trained on the training data, it is now tested on the evaluation data. By introducing new data into the model, the effectiveness of its predictive abilities may be assessed.
  • Parameter tuning
    After the model has been evaluated, specific test parameters may be fine-tuned to produce better results.
  • Predicting
    The final value of the model is realized, and it is used in real-world settings to make informed predictions based on available data.  
How does machine learning work? | ServiceNow

Supervised learning 

Supervised learning describes a machine learning technique in which an algorithm applies what has been learned from data that has been labeled or classified to new data to predict future events. The system provides targets for outputs after being sufficiently trained. It can also compare output to the correct, intended output to identify errors and modify the model as needed.

Unsupervised learning 

Unsupervised learning is used when the information for training is not classified or labeled. It studies how systems infer functions to describe hidden structures and solutions from unlabeled data. It doesn’t necessarily provide the right output, but it is used to explore data and draw different inferences from datasets to identify any hidden structures or interesting relationships.

Semi-supervised learning

This method falls between supervised and unsupervised learning, as it uses both labeled and unlabeled data. It’s typical to use a smaller amount of labeled data and a larger amount of unlabeled data—the systems that apply this method considerably improve learning accuracy. Semi-supervised learning is usually chosen when labeling data requires skilled and relevant resources for training/learning.

Reinforcement learning 

A method that interacts with the environment by producing actions to identify errors or rewards. The key characteristics of reinforcement learning are trial and error search and delayed reward. Simple feedback is required to learn which action is best—this is the reinforcement signal. This allows software agents and machines to determine ideal behavior within a context to maximize performance.

Financial services 

Businesses in the financial services use machine learning technology to identify insights in data and prevent fraud. The insights help locate opportunities for investment. Data mining and machine learning can also identify high-risk clients or utilize cybersurveillance to uncover fraud.


Public safety and utilities can use machine learning, as they have many sources of data to be mined for insights. For instance, they can analyze sensor data to identify ways to save money, detect anomalies, and increase efficiency. Machine learning also helps identify fraud to help minimize identity theft.

Health care 

There is a growing trend of the use of machine learning with the help of wearable devices and sensors that use data to assess the health of a patient in real time or extract the most important information pertaining to patients’ health. This technology can help medical experts analyze and identify trends in data or any issues that can lead to improved treatment and diagnosis.


Websites have the ability, using machine learning, to recommend items that customers may want based on previous purchases and the purchases of others. Retailers capture data, analyze it, and personalize the shopping history, especially with marketing campaigns, price optimization, supply planning, inventory management, and customer insights.

Oil and gas

Machine learning is used to find new energy sources, analyze minerals in the ground, streamline distribution, predict refinery and sensor failure, and other cost-effective movements.


Transportation benefits from making routes more efficient, by analyzing data to identify patterns and trends and predict potential issues for the increase of profitability. Data analysis and modeling aspects of machine learning are crucial for delivery companies and public transportation.

Digital assistants and chatbots

Machine learning may be applied to chatbots and digital assistants, allowing them to evolve and learn from inputs, and working to maintain natural language processing while gathering and storing relevant information.


The use of machine learning for recommendations encompasses everything from streaming services to retail. A machine learning system gathers information about customers over time, draws correlations about the consistent behaviors and patterns, then provides customized recommendations based on the patterns it has ascertained.

Contextual online advertising

Consumers want to see ads that are relevant to them. Machine learning technology helps populate relevant keywords that are trending in content, while also helping marketers take advantage of brand-building content.


An essential aspect of AI security, machine learning helps make cybersecurity simpler, less expensive, more effective, and more proactive. Security AIOps and security operations use machine learning to analyze patterns to help predict and prevent attacks, both similar and new, while adapting to changing behavior.

ServiceNow, the industry leader in business IT solutions, delivers machine learning benefits to organizations across industries. Now Intelligence, powered by the Now Platform, uses machine learning for predictive intelligence. Automate end-to-end workflows, run intelligent operations, identify issues, reduce call volumes, automate solutions to common requests, and identify the patterns that matter most to improving your business—machine learning with ServiceNow makes it all possible.

Get started with the Now Platform 

The Now Platform includes core capabilities that enable you to quickly and efficiently digitize workflows and run them at scale.