When our target variable is numeric, we use regression instead of classification. For this exercise, we will use Auto MPG data set from the Datasets widget. This is a small data set, where the aim is to predict fuel consumption (in miles per gallon, mpg) from 9 variables.

Find the most informative projection in a Scatter Plot. What does it tell you? Are variables correlated? How do they affect fuel consumption?

Use Linear Regression and pass it, along with the data, to Predictions. Now for some data wrangling. Use Feature Constructor to create a new numeric feature. Give it a name (i.e. error). Then compute the difference between the predicted and actual target value (enter Linear_Regression-mpg and press Enter). Connect Scatter Plot to Feature Constructor, find the most informative projection from the previous task and color points with the new feature (i.e. error). What does the plot show? Which cars had the biggest negative error and the biggest positive error?

Which variable is negatively correlated with the fuel consumption? And which one is positively correlated? What is the conclusion? What kind of a car should I buy if I want to have minimum fuel consumption?

Last modified: Monday, 14 March 2022, 11:27 AM