Introduction to Supervised Learning

What is Supervised Machine Learning?

What will you think when you see the term supervised learning; the same logic is applied here too. In supervised learning, there will be a supervisor while the machine learns.

Supervised learning means we have to provide a sample data called training data for train the machine. This train data contains the correct answer. After training the machine, we provide the new data that we need to test. The machine will check the new data with the training data we provided earlier and predict the output.

For better understanding, consider an example that we have a basket of some vegetables, which we need to classify.

Data collection and inputting data

First, we are taking the sample vegetables from the basket. 

Suppose we take a carrot and we feed the characteristics of carrot to machine-like long, tapering towards front and orange in color. 

Next, we take a tomato then we feed it to the machine-like it's red in color and pulpy and round in shape.

Then we are providing the vegetables to the machine and the machine correctly identifies the carrot and tomato with the help of the details we feed to the machine.

The details we first feed to the machine are called Train data and the process is called Training the machine. Then we give input data is called Test data and the machine gives prediction on basis of Supervised learning.

In supervised learning, we are providing the input and the output and machine aiming to find some mapping relation to map the input variable with the output.

In our world supervised learning is used in various fields like fraud detection, email filter, image classification, etc.

How the supervised learning works

As we know supervised learning involves the training of models, using some sample data called training data. In this training, models will learn about the data in the dataset and they can analyze and predict the output of the new data we will provide.

Suppose we have some objects having different shapes and colors.

Data collection and inputting data

First, we have to make sample data from the dataset called training data to train the model. In our case, we are extracting the objects of round, rectangle, and pentagon into train data

Then we train the model for the data as

  1. If the object is round in shape with red color then we classify it as a red ball
  2. If the object is a rectangle (four sides) in shape and green color we classify it as a green block
  3. If the object is a pentagon(five sides) and white in color we classify it as a white pentagon

Now the model trained using the train data and then we provide the new data (new objects) to the model so that the model can easily classify them according to the trained data(color and number of sides) and predict the output.

Steps for Supervised learning

Data collection and inputting data
  1. Accept the dataset that wants to analyze
  2. Collect the sample data called train data
  3. Split the training dataset into train dataset, validation set, and test set.
  4. Check the features of the train dataset so that it can predict the output for all the datasets. 
  5. Select the proper machine learning algorithm for the model depends upon the output data and input data.
  6. Apply the train data to the algorithm to train the model and check it using the validation set.
  7. Check the model using the test set before providing the new data to check the efficiency

Advantages of supervised learning

  • Models are predicting the output with solid experience so more accurate.
  • We have much more knowledge about the dataset and the objects
  • Help to increase performance with the help of training models
  • Help to solve complex problems in computation.

Disadvantages of Supervised learning

  • It cannot be much reliable in solving complex tasks.
  • Training the model using sample data needs a lot of time
  • If we provide any data, which is not in the training data, it will give an error.
  • We need much knowledge about the dataset and the classes in them.
  • Classifying a huge amount of data is much hard.

Types of Supervised machine learning algorithms

Supervised Learning

Supervised machine learning algorithms can be further divided into two depends upon the task we have to complete with the algorithm. that are

  1. Regression
  2. Classification

Regression

Regression algorithm is used for weather forecast, stock predictions, etc because regression algorithm needs a relation between the input and output variables. In regression, the output must be a real value and it predicts only a single output using the train data.
For example, if you want to predict a stock value and the input will be many factors like company data, performance, etc. but the output will be some real single value of stock amount.
There are many algorithms that work under the regression. Some of them are,

  1. Linear regression
  2. Logistic regression
  3. Regression trees
  4. Nonlinear regression
  5. Bayesian linear regression
  6. Polynomial regression

Classification

As the name suggests, this algorithm is used to classify or categorize the given dataset into different classes. This means it is used when the output has some categorical data in it.
If the data is classified into only two types, For example, classify as men and woman, true or false, animal or bird-like that it is called binary classification. 
If it classified into more than two classes then it is multiclass. For example, we have to classify fruits into different classes. There are many algorithms that work under classification and they are

  1. Naïve Bayes classifiers
  2. Random Forest
  3. Decision trees
  4. Support vector machines