Regresyon
The task of monitoring learning that estimates a continuous numerical value that corresponds to an input.
Regretion is an audited learning task aimed at predicting a numeric value not constantly (continuous) that corresponds to an input. The main difference from classification is that the output is not a final set of categories, theoretically it comes from a number axis that can be called endless numbers; for example, the price of a house is a regression problem to predict the temperature of the half or the duration of the stay in the hospital. Linear regression, which is the simplest state, tries to find a right that represents this relationship best by having a linear relationship between input variables and output; polynom regression for more complex relationships, decision-based regression models or deep nerve networks can be used.
Regretion models are often trained to minimize the difference between predicted value and actual value, for this purpose, lost functions such as average square error (MSE) or an average absolute error (MAE). Regretion is widely used in stock and sales forecasts in the financial sector, price value in real estate, weather forecast in meteorology, predicting the progress speed of the disease in health and modeling of continuous physical size, such as material durability in engineering; creates one of the most basic two types of problems with the classification of machine learning.
Regretion models are often trained to minimize the difference between predicted value and actual value, for this purpose, lost functions such as average square error (MSE) or an average absolute error (MAE). Regretion is widely used in stock and sales forecasts in the financial sector, price value in real estate, weather forecast in meteorology, predicting the progress speed of the disease in health and modeling of continuous physical size, such as material durability in engineering; creates one of the most basic two types of problems with the classification of machine learning.