40 Algorithms Every Programmer Should Know

The problem statement

Given the historical data, can we train a binary classifier that can predict whether a particular user will eventually buy a product based on their profile?

First, let's explore the historical labeled data set available to solve this problem:

x € ℜ^b, y € {0,1}

For a particular example, when y = 1, we call it a positive class and when y = 0, we call it a negative class.

Although the level of the positive and negative class can be chosen arbitrarily, it is good practice to define the positive class as the event of interest. If we are trying to flag the fraudulent transaction for a bank, then the positive class (that is, y = 1 ) should be the fraudulent transaction, not the other way around.

Now, let's look at the following: ...