Problem Definition (1/6).
Problem Definition is the first step to take when building a Machine Learning (ML) model. It is a concise description of the problem to be solved. It is important to define and know the problem you are trying to solve as this reduces the amount of time used to tackle the problem. It also gives a hint on how to solve the problem and the tools to use. There are 4 major types of Machine Learning depending on the type of problem. These are:
1. Supervised Learning.
Supervised Learning is a type of Machine Learning that has labels in the data. This type of learning is always organized and easier to read and view. The main types of Supervised Learning are:
Classification: This is a method of Supervised Learning that defines the class to which data elements belong and is best used when finite and discrete values are available for the output. It also forecasts a class for an input variable. An example is the prediction of cancer in a patient.
Regression: This is a type of Supervised Learning that predicts a continuous outcome variable (y) based on the value of one or multiple predictor variables (x). For example, the prediction of house prices in a state.
2. Unsupervised Learning.
This is the opposite of Supervised Learning. This kind of learning is neither organized nor labeled i.e. it makes use of unstructured data. Examples of the kind of data used are images, audio files, and videos. A popular type of Unsupervised Learning is Clustering.
Clustering is simply a list of data based on their similarity and dissimilarity. From the name, it means grouping data in clusters to get the best fit. One example of this is clustering users for movie recommendations.
3. Transfer Learning.
Transfer learning is the transfer of knowledge from one ML model to another. It is a type of learning that focuses on the preservation of information acquired while solving one issue and applying it to a separate but related issue. An example is using the same knowledge used to recognize cars to recognize trucks.
4. Reinforcement Learning.
Reinforcement Learning is the preparation of machine learning models to make a series of decisions. The ultimate goal is to maximize the total reward of the system. This type of learning uses trial and error to come up with solutions. The system offers punishments for the action it performs in order to get what is needed from it.
Depending on what is to be achieved, the various learning is usually used for different purposes. It is best to understand the problem before selecting what model to employ.
You can read the main article here :)