What are types of machine learning?
What is supervised learning?
What is Learning for a machine?
Lets define it in terms of “Past Experience”, “Task” and “Performance”.A machine is said to be learning the things from past Experiences(input data feed in) w.r.t some class of Tasks, if it’s Performance in a given Task progresses along with the Experience.
Supervised Learning :
Supervised Learning is the task of learning a function that map the input to output pair based on the example of input-output pair.
In this type of learning process, model is getting trained on a dataset called labelled dataset. Labelled dataset is that type of dataset which have both input and output parameters. Training and validation datasets are labelled in the below figures:
Above figures have labelled data set –
- Figure A: Here in this figure we can find dataset of a shopping store. It predicts whether a customer will purchase a particular product under consideration or not. And this prediction is based on his/ her gender, age and salary.
Input Parameter : Gender, Age, Salary,Address
Output Parameter: Purchased i.e. 0 or 1 ; the meaning of 1 is yes the customer will purchase and 0 means that customer won’t purchase it.
- Figure B: Figure B shows the Meteorological dataset which serves the purpose of predicting wind speed based on different parameters such as Dew point, Temperature, Pressure, relative Humidity, Wind Direction.
Input Parameter : Dew Point, Wind Direction, Temperature, Pressure, Relative Humidity,
Output Parameter : Wind Speed
Training the system:
When we train the model, the dataset is split in the ratio of 80:20 i.e. 80% is training data and rest 20% is testing data.
In training data phase, we feed input as well as output for 80% data. The model learns from training datasets only. We use various machine learning algorithms which we will discuss in the next chapter to build our model. When the model learns, it will build some logic of its own.
After the model is ready, it is good to be tested. When testing the model, input is fed from remaining 20% data which the model has never seen before. The model will predict some value of data and compare it with actual output and calculate the accuracy.
Types of Supervised Learning:
Classification is a Supervised Learning task where output is having defined discrete value called label. In the above Figure A, Output – Purchased data has defined labels of either 1 or 0 ; 1 means the customer will purchase the product where as 0 means that customer won’t purchase the product. The main goal here goal is to predict discrete values belonging to a specific class and evaluate on the basis of accuracy.
Classification is of two types:
- Binary classification
- Multi class classification.
In binary classification, it will predicts either 0 or 1 ; yes or no
Where as in multi class classification, model predicts more than one class.
Example: Gmail categorizes mails in more than one classes like social, promotions, updates.
Regression :Regression is a type of Supervised learning where output is having continuous value.
In the above Figure B, Output – Wind Speed is not having any discrete value . The output values is continuous in the particular range. The main task here is to predict a value as much closer to actual output value. After that evaluation is done by calculating error value. The small the error means the greater the accuracy of our regression model.
Supervised learning example:
- Linear Regression
- Nearest Neighbor
- Support Vector Machine (SVM)
- Random Forest
- Guassian Naive Bayes
- Decision Trees
Linear Regression: It is a type of machine learning algorithm which allows to map numeric inputs to numeric outputs that fits a line into the data points.
Logistic Regression: A type of classification algorithm which is widely used when the dependent variable is binary i.e 0 or 1.
Neural Networks: A type of Machine Learning framework that gets its effectiveness as introduced from non-linearity to linear ML models
Support Vector Machines: A type of Machine Learning algorithm which uses Margin Maximization in determining and developing the optimal separator line between classes, using the Kernel Trick.
These algorithm can be used in:
- Image classification
- Speech recognition
3. Regression-based number prediction