Chapter 6. Machine Learning

Machine learning is a broad topic with many different supporting algorithms. It is generally concerned with developing techniques that allow applications to learn without having to be explicitly programmed to solve a problem. Typically, a model is built to solve a class of problems and then is trained using sample data from the problem domain. In this chapter, we will address a few of the more common problems and models used in data science.

Many of these techniques use training data to teach a model. The data consists of various representative elements of the problem space. Once the model has been trained, it is tested and evaluated using testing data. The model is then used with input data to make predictions.

For example, the purchases made by customers of a store can be used to train a model. Subsequently, predictions can be made about customers with similar characteristics. Due to the ability to predict customer behavior, it is possible to offer special deals...

Supervised learning techniques

There are a large number of supervised machine learning algorithms available. We will examine three of them: decision trees, support vector machines, and Bayesian networks. They all use annotated datasets that contain attributes and a correct response. Typically, a training and a testing dataset is used.

We start with a discussion of decision trees.

Decision trees

A machine learning decision tree is a model used to make predictions. It effectively maps certain observations to conclusions about a target. The term tree comes from the branches that reflect different states or values. The leaves of a tree represent results and the branches represent features that lead to the results. In data mining, a decision tree is a description of data used for classification. For example, we can use a decision tree to determine whether an individual is likely to buy an item based on certain attributes such as income level and postal code.

We want to create a decision tree that...

Unsupervised machine learning

Unsupervised machine learning does not use annotated data; that is, the dataset does to contain anticipated results. While there are several unsupervised learning algorithms, we will demonstrate the use of association rule learning to illustrate this learning approach.

Association rule learning

Association rule learning is a technique that identifies relationships between data items. It is part of what is called market basket analysis. When a shopper makes purchases, these purchases are likely to consist of more than one item, and when it does, there are certain items that tend to be bought together. Association rule learning is one approach for identifying these related items. When an association is found, a rule can be formulated for it.

For example, if a customer buys diapers and lotion, they are also likely to buy baby wipes. An analysis can find these associations and a rule stating the observation can be formed. The rule would be expressed as {diapers, lotion...

Reinforcement learning

Reinforcement learning is a type of learning at the cutting edge of current research into neural networks and machine learning. Unlike unsupervised and supervised learning, reinforcement learning makes decisions based upon the results of an action. It is a goal-oriented learning process, similar to that used by many parents and teachers across the world. We teach children to study and perform well on tests so that they receive high grades as a reward. Likewise, reinforcement learning can be used to teach machines to make choices that will result in the highest reward.

There are four main components to reinforcement learning: the actor or agent, the state or scenario, the chosen action, and the reward. The actor is the object or vehicle making the decisions within the application. The state is the world the actor exists within. Any decision the actor makes occurs within the parameters of the state. The action is simply the choice the actor makes when given a set of options...

Summary

Machine learning is concerned with developing techniques that allow the applications to learn without having to be explicitly programmed to solve a problem. This flexibility allows such applications to be used in more varied settings with little to no modifications.

We saw how training data is used to create a model. Once the model has been trained, the model is evaluated using testing data. Both the training data and testing data come from the problem domain. Once trained, the model is used with other input data to make predictions.

We learned how the Weka Java API is used to create decision trees. This tree consists of internal nodes that represent different attributes of the problem. The leaves of the tree represent results. Since there are many ways of constructing a tree, part of the job of a decision tree is to create the best tree.

Support vector machines divide a dataset into sections thus classifying elements in the dataset. This classification s based on the attributes of...