Machine learning algorithms are used primarily for the following types of output:
- Clustering (Unsupervised)
- Two-class and multi-class classification (Supervised)
- Regression: Univariate, Multivariate, etc. (Supervised)
- Anomaly detection (Unsupervised and Supervised)
- Recommendation systems (aka recommendation engine)
The specific methods utilized for each output type will be explained in the next section, but first, let's have a look at each of the above output, or problem categories in general.
Clustering, as previously stated, is an unsupervised approach for determining the composition and structure of a set of data. It's a method of clumping data into clusters to observe what, if any, groups emerge. A cluster centroid and a confined collection of data points characterize each cluster. The cluster centroid is essentially the mean (average) of all the data points in the cluster, regardless of characteristic.
Placing a data point (also known as an observation) into a pre-defined class or category is a classification issue. Classification challenges can range from simply assigning a class to an observation to estimating the likelihood that an observation belongs to each of the specified classes.
Assigning the class of Spam or Ham to an incoming email, where ham simply means 'not spam,' is a fantastic example of a two-class categorization. More than two potential classifications are referred to as multi-class classification. So, in the spam example, 'Unknown' may be a third class.
The term "regression" simply means that a model will give a continuous value (response) to a data observation rather than a discrete class. Predicting the closing price of the Dow Jones Industrial Average on any given day is a perfect illustration of this. This value might be any number, making it an excellent choice for regression analysis.
It's worth noting that the term "regression" is occasionally used to refer to an algorithm that's really employed to solve classification difficulties or anticipate a discrete categorical answer (e.g., spam or ham). Logistic regression, for example, predicts the probability of a discrete value.
Anomaly detection is another issue category. While we'd like to believe that data is well-behaved and rational, this isn't always the case. Erroneous data points can occur as a result of malfunctions or measurement mistakes, as well as fraud. Anomaly readings may also be suggestive of a malfunctioning piece of hardware or electronics.
When anomalies are symptomatic of a real problem that is difficult to explain, such as a manufacturing fault, spotting anomalies gives a measure of quality control as well as insight into whether or not efforts made to eliminate defects have worked. In either case, there are occasions when finding these anomalous values is advantageous, and certain machine learning methods may be utilised to accomplish so.
A recommendation system, often known as a recommendation engine, is used to solve the last kind of problem. Recommendation systems are a sort of information filtering system used to offer suggestions in a variety of applications such as movies, music, books, restaurants, articles, and items. Content-based and collaborative filtering are the two most frequent ways.
Summary
Machine learning, predictive analytics, and other related subjects are fascinating and powerful.
While these subjects might be somewhat complex, many of the principles are fairly straightforward to grasp at a high level. In many circumstances, all that is necessary for talks about machine learning challenges, projects, methodologies, and so on is a basic comprehension. If you are interested in learning all the concepts of Machine Learning then Consider taking Machine Learning Certification Training.
Top comments (0)