We've seen from the previous examples, that we need features, such as whether a fruit can be green, yellow, or red, or whether it's tropical. We're now focused on the project at hand. What should the features be?:
|
Class |
??? |
??? |
??? |
|
Spam |
|||
|
Ham |
What makes up an email? Words make an email. So, it would be appropriate to consider the appearance of each word feature. We can take it further, and take the intuition that we have developed previously with TF-IDF and instead use the frequency of the words among the document types. Instead of counting 1 for the existence, we count the total number of times a word exists in the document types.
The table would look something as follows:
|
Class |
Has XXX |
Has ... |
