Bloc 4 : [ Analyse prédictive de données non-structurées par l'intelligence artificielle 📁🔍 ] Mandatory for JedhaBootcamp certification

AT&T Orange SMS Spam Detector🕵️‍♀️

One of the main pain point that AT&T users are facing is constant exposure to SPAM messages. AT&T has been able to manually flag spam messages for a time, but they are looking for an automated way of detecting spams to protect their users.

Mail : fezzibasma@gmail.com

Company's Description 📇

AT&T Inc. is an American multinational telecommunications holding company headquartered at Whitacre Tower in Downtown Dallas, Texas. It is the world's largest telecommunications company by revenue and the third largest provider of mobile telephone services in the U.S. As of 2022, AT&T was ranked 13th on the Fortune 500 rankings of the largest United States corporations, with revenues of $168.8 billion! 😮

Project 🚧

One of the main pain point that AT&T users are facing is constant exposure to SPAM messages.

AT&T has been able to manually flag spam messages for a time, but they are looking for an automated way of detecting spams to protect their users.

Goals 🎯

Your goal is to build a spam detector, that can automatically flag spams as they come based sollely on the sms' content. Write a notebook that runs preprocessing and trains one or more deep learning models in order to predict the spam or ham nature of the sms State the achieved performance clearly

Result

Sequential Simple Model ( training less than 1 min for 50 epoches )

Transfer Learning using BERT Model ( training for 5 epoches = 2 hours )

Credits hub.KerasLayer to import BERT pretained model : https://www.tensorflow.org/hub/api_docs/python/hub/KerasLayer

What about performance ?

The easiest model "Sequential Simple Model" is fast to train and got better results !