Code With Me : Building a Spam Filter !

Ever wonder how a spam filter works? Let's code one together!
Bag of Words Video: • Bag of Words : Natural...
Link to Code : github.com/ritvikmath/KZbin...
Link to Data : www.kaggle.com/uciml/sms-spam...
My Patreon : www.patreon.com/user?u=49277905

[@tamojitmaiti]

Have been consistently watching your videos for a while now. It's amazing how clear, concise and succinctly you manage to explain everything. Just a suggestion, it'd be massively helpful if you could link/mention relevant books pertaining to topics in the video description. In the age of SEO laden blog posts, filtering out noise amidst Google searches is a pain.

[@drsandeepvm5622]

Thanks for your exceptionally simple explanation 👍

[@gracikk]

Really good explanation. Thank you!

[@kdhlkjhdlk]

In my experience, using word embedding (fasttext is good for the character ngrams) + gradient boosting is far better than naive bayes.

[@photonicsauce7729]

This is just naive bayes maybe. Cool

[@YohaneesHutagalung]

Could you make a video about stock predictions with RNN Lstm

[@Blaze098890]

spoiler: it doesn't really work

[@ccuuttww]

LOL I think either your dataset or something went wrong in most situation Naive Bayes always has around 93% accuracy even u do the following technique lemmize steeming remove stopword N-Grams but anyway this video give us a concept of how it works with python

[@Islamic_Tv984]

link please

[@jessicatriplev9802]

I think you've made a mistake in: test_spam_df = spam_df.iloc[int(len(spam_df)*0.7):] The testing set should have been 30% not 70% of the data. That's perhaps why the validation result was so good.

[@gracikk]

test_spam_df = spam_df.iloc[int(len(spam_df)*0.7):] It basically means that we save in our test set only those observations which index is more or equal to len(spam_df)*0.7

[@jessicatriplev9802]

@@gracikk Right. But that means that 70% of the data will end up in the testing set. He must have meant 30%.

[@gracikk]

​@@jessicatriplev9802I can't send you a colab link with a simple example, youtube delete it. But you can try it on your own to be sure that everything is fine

[@joshuasigelman8141]

I think it's fine because he is taking the first slice UNTIL the 70% mark and then the testing set FROM the 70% mark.

[@GeyzsonKristoffer]

Bypass the model by repeating words that are biased towards the non-spam, multiple times.

[@BurkenProductions]

Wrong language for this. Need php

[@Lizergus]

lmao