Machine Learning Regression

Linear regression algorithm predicts continous values (like price, temperature). This is another article in the
algorithms for beginners series. It is a supervised learning algorithm, you need to collect training data for it to work.

Linear Regression

Introduction

Classification output can only be discrete values. There can be [0],[1],[2] etcetera. What if you want to output prices or other continous values?

Then you use a regression algorithm.

Lets say you want to predict the housing price based on features. Collecting data is the first step. Features could be number of rooms, area in m^2, neighborhood quality and others.

Example

Write down the feature: #area_m2. For our example in code that looks like this.

from sklearn.linear_model import LinearRegression

X = [[4], [8], [12], [16], [18]]
y = [[40000], [80000], [100000], [120000], [150000]]

model = LinearRegression()
model.fit(X,y)

# predict
rooms = 11
prediction = model.predict([[rooms]])
print('Price prediction: $%.2f' % prediction)

Then you can create a plot based on that data (if you want to). You see there is a correlation between the area and the price.

This is a linear relationship. You can predict the price, with a linear regression algorithm.

Explanation

First you import the linear regression algorithm from like it learn then you defined a training data X and the Y where axis the area and y is the price.

model = LinearRegression()
model.fit(X,y)

Linear regression algorithm because there is a linear relationship then we train the algorithm using the training data.

Now that the algorithm is trained you can make predictions using the area. A new example, can predict the price for you.

rooms = 11
prediction = model.predict([[rooms]])
print('Price prediction: $%.2f' % prediction)

This algorithm LinearRegression only works if there is a linear relation in your data set. If there isn't, you need a polynomial algorithm.

Plot to verify that there is a linear relation.

Download examples and exercises