MATLAB Tools for Recommender Systems

A recommender system is a model that’s designed to anticipate the preferences of a specific user. When applied to the domain of movies, it transforms into a movie recommendation engine. The process involves filtering items in a database by predicting the user’s potential ratings and facilitating the connection of users with the most suitable content in the dataset. This holds significance because, in extensive catalogs, users might not discover all pertinent content. Effective recommendations enhance content consumption and major platforms such as Netflix heavily depend on them to maintain user engagement. In this chapter, we will learn the basic concepts of recommender systems and how to build a network intrusion detection system (NIDS) using MATLAB.

In this chapter, we’re going to cover the following main topics:

• Introducing the basic concepts of recommender systems
• Finding similar users in data
• Creating...

Technical requirements

In this chapter, we will introduce basic machine learning concepts. To understand these topics, a basic knowledge of algebra and mathematical modeling is needed. You will also required working knowledge of MATLAB.

To work with the MATLAB code in this chapter, you’ll need the following files (available on GitHub at https://github.com/PacktPublishing/MATLAB-for-Machine-Learning-second-edition):

• CreditCardData.xlsx
• CreditCardFraudDet.m
• NDISdata.csv
• NDISEnsemble.m

Introducing the basic concepts of recommender systems

A recommender system is a type of information filtering system that’s designed to suggest items or content to users based on their preferences, historical behavior, or other relevant factors. These systems are widely used in various online platforms to help users discover products, services, content, and more. Recommender systems involve two primary entities: users and items. Users are individuals for whom recommendations are generated, and items are the products, content, or services to be recommended. These items can include movies, books, products, news articles, and more.

Recommender systems rely on data that captures the interaction between users and items. This interaction data can include user ratings, purchase history, clicks, views, likes, and any other form of user engagement with items.

There are different types of recommender systems:

• Collaborative filtering (CF): CF methods make recommendations...

Finding similar users in data

Fraud has consistently been a pervasive issue in various forms, but the emergence of new technological tools, such as virtual intelligence (VI), has expanded the avenues for fraudulent activities. In today’s world, the use of credit and debit cards has become the standard for making purchases, and as a result, fraud associated with these payment methods is on the rise. The repercussions of such fraud extend beyond impacting just merchants and banks, who are often left shouldering the financial burden.

When a customer falls victim to fraud, they may find themselves burdened with higher interest rates imposed by the bank as they could be categorized as a higher risk profile. Additionally, fraud incidents can tarnish a merchant’s reputation and image. If a customer experiences fraud during a transaction, it can erode their trust in the seller, potentially driving them to seek alternatives from competitors for future purchases.

Given a...

Creating recommender systems for network intrusion detection using MATLAB

A NIDS serves as a security mechanism that’s employed to identify and prevent unauthorized access, malicious activities, and potential threats within a computer network. It involves monitoring network traffic and analyzing it to identify any suspicious or abnormal behaviors. The main objective of network intrusion detection is to protect the network from various types of attacks, such as denial-of-service (DoS) attacks, malware infections, data leakage, unauthorized access, and other cyber threats.

There are two primary methods of network intrusion detection:

• Signature-based detection: This method involves comparing network traffic patterns with a database of known signatures or patterns of known attacks. If a match is found, an alert is generated to notify the network administrator.
• Anomaly-based detection: This method focuses on identifying abnormal or suspicious network behavior that...

Deploying machine learning models

Deploying machine learning models refers to the process of making a trained model available for making predictions on new, unseen data. It involves taking the trained model and integrating it into a production environment where it can receive input data, perform predictions, and return the results. The trained model needs to be organized and packaged into a format suitable for deployment. This may involve exporting the model into a file format that can be easily loaded and used by other systems. An application programming interface (API) is typically created to expose the machine learning model’s functionality. The API acts as the interface that other systems or applications can use to send data and receive predictions from the model.

If the model is expected to handle many concurrent requests, the deployment environment may need to be scaled to accommodate the increased load. This may involve setting up clusters of servers or using cloud...

Summary

In this chapter, we learned the basic concepts of recommender systems, starting with the definition of these systems and then understanding how the problem is approached. We analyzed the different types of recommender systems: CF, content-based filtering, and hybrid recommender systems.

Next, we saw how to use similarities in the data to identify possible fraudulent uses of credit cards. To do this, we trained a model based on the nearest neighbor algorithm but using a modified version of the traditional k-NN algorithm, where neighbors are given varying weights during the prediction or classification process.

Then, we saw how to implement a NIDS based on ensemble methods in MATLAB. Specifically, we adopted an AdaBoost algorithm to identify intrusions in a LAN network.

Finally, we introduced the techniques of deploying machine learning models regarding model compression. We analyzed the most popular model compression techniques, including pruning, quantization, knowledge...