Modelling of a Deep Learning Based SMS Spam Detection Application

Abstract

Various techniques for mitigating spam messages have been
developed overtime. Methods that involve spam emails have also been
applied to Short Message Service domain. Some of these methods are
based on Artificial Intelligence. This work is aimed at the modelling
and deployment of a spam detection application using Deep Learning
algorithms. Spam data repositories were investigated for appropriate
data required to create the model. Packages such as Google Colab,
Pandas, Seaborn, Matplotlib and Wordcloud were implemented for
expository data analysis to gain insight into the data. Packages such
as Tensorflow, Python-dotenv, Scikit learn were used to create and
evaluate the Deep Learning model. The model was thereafter deployed
using Flask, a python library for web development. The web
application was used to read SMS messages sent to a Twilio service
phone number and messages as spam or ham. The results show an
accuracy score of 98.3% hence the developed model is highly reliable.