What is machine learning and why is it important? | micro focus (2023)

Machine learning is a subset ofartificial intelligencefocuses on building systems that can learn from historical data, recognize patterns, and make logical decisions with little or no human intervention. It is a data analysis method that automates the construction of analytical models using data that includes various forms of digital information, including numbers, words, clicks, and images.

Machine learning applications learn from input data and continuously improve the accuracy of the results through automated optimization methods. The quality of a machine learning model depends on two main aspects:

1.The quality of the input data.A common phrase when developing machine learning algorithms is "garbage in, garbage out". The saying means that if you input poor quality or garbled data, the output of your model will be quite inaccurate.

2.The choice of the model itself.In machine learning, there are a variety of algorithms that a data scientist can choose from, each with their own specific uses. It is important to choose the right algorithm for each use case. Neural networks are a type of algorithm with great expectations due to the great precision and versatility they can offer. However, for small data sets, choosing a simpler model generally works better.

The better the machine learning model, the more accurately it can find features and patterns in the data. This, in turn, means your decisions and forecasts will be more accurate.

Accelerate your SOC with machine learning

Why is machine learning important?

Why Use Machine Learning? Machine learning is becoming increasingly important due to the increasing volume and variety of data, the accessibility and affordability of computing power, and the availability of high-speed internet. Thisdigital TransformationFactors enable the rapid and automated development of models that can quickly and accurately analyze extraordinarily large and complex data sets.

There are a large number of use cases where machine learning can be applied to reduce costs, mitigate risk and improve overall quality of life, including product/service recommendations,Cyber ​​security breach detectionand empowerment of autonomous vehicles. As access to data and computing power increases, machine learning is becoming more ubiquitous and will soon be integrated into many facets of human life.

How does machine learning work?

There are four main steps to follow when creating a machine learning model.

1. Choose and prepare a training data set

Training data is information representative of the data that the machine learning application feeds to adjust model parameters. Training data is sometimes labeled, meaning it has been labeled to indicate rankings or expected values ​​that machine learning mode should predict. Other training data may not be labeled, requiring the model to extract features and assign clusters autonomously.

For labeling, the data must be split into a training subset and a test subset. The former is used to train the model and the latter is used to evaluate the model's effectiveness and find ways to improve it.

2. Choose an algorithm to apply to the training data set

The type of machine learning algorithm you choose mainly depends on a few things:

• Whether the use case is predicting a value or order using labeled training data or whether the use case is clustering or dimension reduction using unlabeled training data.
• How much data is in the training set
• The type of problem the model is trying to solve.

For prediction or classification use cases, you would typically use regression algorithms such as general least squares or logistic regression. For unlabeled data, you probably rely on clustering algorithms like k-means or nearest neighbor. Some algorithms, such as B. neural networks can be configured to work with clustering and prediction use cases.

3. Train the algorithm to build the model

Algorithm training is the process of tuning model variables and parameters to more accurately predict reasonable outcomes. The training of machine learning algorithms is usually iterative and uses different optimization methods depending on the model chosen. These optimization methods do not require human intervention, which is part of the power of machine learning. The machine learns from the data you feed it with little or no specific direction from the user.

4. Use and improve the model

The last step is to feed the model with new data to improve its effectiveness and accuracy over time. The source of new information depends on the nature of the problem to be solved. For example, a machine learning model for autonomous cars will contain real-world information about road conditions, objects, and traffic rules.

Use cases for machine learning in the SOC

Machine learning methods

What is Supervised Machine Learning?

supervised machine learningThe algorithms use data, called training data, where the corresponding outputs for the input data are known. The machine learning algorithm takes a set of correct inputs and corresponding outputs. The algorithm compares its own predicted results with the correct results to calculate the accuracy of the model, and then optimizes the model parameters to improve accuracy.

Supervised machine learning relies on patterns to predict values ​​in unlabeled data. It is most commonly used in automation, with large amounts of data sets, or in cases where too much data is being input for a human to process efficiently. For example, the algorithm can identify credit card transactions that are likely to be fraudulent, or identify the insurance customer who is likely to make a claim.

What is unsupervised machine learning?

unsupervised machine learningIt is best applied to data that does not have a structured or objective answer. There is no default value for the correct output for a given input. Rather, the algorithm must understand the input and make the appropriate decision. The aim isExamine the information and identify the structure within it.

Unsupervised machine learning works well with transaction information. For example, the algorithm can identify customer segments that share similar attributes. Customers in these segments may be the target of similar marketing campaigns. Popular techniques used in unsupervised learning include nearest neighbor mapping, self-organizing maps, singular value decomposition, and k-means clustering. The algorithms are then used to segment topics, identify outliers, and recommend articles.

What is the difference between supervised and unsupervised machine learning?

What machine learning can do: Machine learning in the real world

While machine learning has existed for decades, it's the more recent ability to automatically apply and compute complex mathematical calculations using big data that have given it unprecedented sophistication. The domain of machine learning applications today is vast, by enterpriseAIOpsto online trading. Some real-world examples of machine learning resources today include the following:

• Internet securityUse behavior analysis to identify suspicious or abnormal events that may indicate thisinternal threats,SUITABLEor zero-day attacks.
• Autonomous vehicle projects such asWaymo(a subsidiary of Alphabet Inc.) and TeslaAutopilotThat's a step below true self-driving cars.
• Digital assistants like Siri, Alexa and Google Assistant that scan the internet for information in response to our voice commands.
• User personalized recommendations powered by machine learning algorithms on websites and apps like Netflix, Amazon and YouTube.
• fraud detection etcCyber ​​Resilience SolutionsAggregate data from multiple systems, discover high-risk customers and identify patterns of suspicious activity. These solutions can use supervised and unsupervised machine learning to classify financial organization transactions as fraudulent or legitimate. Because of this, a consumer may receive text messages from their credit card company confirming that an unusual purchase using the consumer's financial information is legitimate. Machine learning has progressed so far in the field of fraud that many credit card companies advertise guilt-free to consumers when fraudulent transactions are not detected by the financial organization's algorithms.
• Image recognition has made significant advances and can be reliably used for facial recognition, reading handwriting on deposited checks, monitoring traffic, and counting the number of people in a room.
• Spam filters that detect and block unwanted emails from inboxes.
• utilities thatAnalyze sensor dataFinding ways to improve efficiency and reduce costs.
• Wearable medical devices that capture valuable real-time data for use in ongoing patient health assessments.
• Taxi apps evaluate real-time traffic conditions and recommend the most efficient route.
• Sentiment analysis determines the tone of a line of text. Good uses of sentiment analysis are Twitter, the opinions of customers and respondents:
  • Twitter: One way to evaluate brands is by recognizing the tone of tweets aimed at a person or business. Companies like Crimson Hexagon and Nuvi offer this real-time.
  • Customer Reviews: You can hear the tone of customer reviews to measure your business performance. This is especially useful if you don't have a rating system combined with free text customer reviews.
  • Polls – Using sentiment analysis on free-text poll responses can provide a quick assessment of how respondents are feeling. Qualtrics has implemented this with its surveys.
• Market segmentation analysis uses unsupervised machine learning to group customers based on their buying habits to determine different types or personalities of customers. This allows you to better understand your most valuable or underserved customers.
• It's easy to press Ctrl+F to search for exact words and phrases in a document, but if you don't know the exact text you're looking for, finding documents can be difficult. Machine learning can use techniques like fuzzy methods, and topic modeling can make this process much easier for you to dosearch for documents without knowing the exact term you are looking for.

Analysis of text and rich media with machine learning technology

The role of machine learning will continue to grow

As data volumes increase, computing power increases, Internet bandwidth increases, and data scientists hone their expertise, machine learning will continue to drive deeper and greater efficiencies at work and at home.

With the increasing cyber threats facing organizations today, machine learning is necessary to protect valuable data and keep hackers off internal networks. Our flagship UEBA SecOps software, ArcSight Intelligence, uses machine learning to detect anomalies that could indicate malicious actions. It has a proven track record of detecting insider threats, zero-day attacks, and even aggressive Red Team attacks. Take the first step in protecting your business by scheduling an ArcSight Intelligence demo today!