Machine Learning models are mainly used for two tasks such as classification task and regression task. A classification problem is about predicting a discrete class label and regression task is prediction a continuous quantity.
Classification
The classification problem can be a binary classification or multi-class classification.
- A problem corresponding two class is called binary classification.
- A problem corresponding to multiple class is called multi-class classification.
Generally, the classification model predicts the probability of each class for a given input. The probability can be interpreted is a likelihood of a given input to each class. The highest probability class can be considered as a predicted class label of a given input.
Example of binary Classification task:
- Predict the email is spam or not spam.
- Predict the image contain dog or cat.
- Predict the voice of female or male.
- predict the person has cancer or not.
Example of multi-class classification task:
- Predict the emotion of a person such as sad, happy, neutral, angry.
- Predict the digit (0-9) from a digit image.
- Predict the image of a dress into a different pattern
- Predict the sentiment of a text.
Regression
Regression predictive is a task of predicting a continuous value of given input variables. A continuous output variable can be an integer or floating-point value.
Example:
- Predict the house price from various house features such as area, the number of bedrooms, built year, etc.
- Predict the share market movements.
- Predict the quantity of rain by analysis of the weather.
. . .
Machine Learning Model
There are various machine learning algorithms exists to solve the classification and regression problems. These model are mainly classified into two categories such as tree-based algorithm and non-tree based algorithm.
Tree-based algorithm:
Tree-based algorithms have a powerful tree-like structure which split the space into sub-space. Tree-based algorithms have a non-linear relationship with its features. They use a divide and conquer approach for splitting. This kind of algorithms have multiple trees and the final outcome is made by combining trees. Below is the list of tree-based algorithms:
- Decision Tree
- Random Forest
- Gradient Boosting Tree
Non-Tree based algorithm:
An Algorithm is called non-tree based algorithm which doesn’t have a tree-like structure. Below is the list of non-tree based algorithms:
- Linear Model: It is a very simple approach. It split space into two subspaces.
-
- Logistic Regression
- Support Vector Machine (SVM)
-
- Neural Networks (NN) : It has a very complex structure. It produces the non-linear decision boundary.
- K-Nearest Neighbour (KNN) : It makes groups of high similarity data points and heavily rely on measurement between points.
. . .