The aim of this project is to create an app which could classify any given message as whether its a SPAM or not.

Irrelevant or unsolicited messages sent over the Internet, typically to a large number of users, for the purposes of advertising, phishing, spreading malware, etc.

• It consumes Internet resources. A deluge of spam will clog mail servers, making all email slow and burdening the ISP.
• It reduces the effectiveness of legitimate advertising.
• It raises costs for everyone who uses the Internet.
• It exposes children to inappropriate material.
• It wastes people's time. This costs the world economy billions of dollars per year in lost productivity.
• It threatens the very utility of email as a form of communication

• The dataset for this project was taken from UCI ML repository. Size of dataset is 5572*2.

• Performed Data preprocessing & Data cleaning on the dataset.

• Performed Text preprocessing (Stemming & lemmatization).

• Applied NLP technique Bag of words on stemmed & lemmatized data individually.

• Applied NLP technique TF-IDF on stemmed & lemmatized data individually.

• Converted the labelled categorical dependent variable into dummy variable to apply ML algorithm.

• Used NaiveBayes classifier model & trained the model using the datasets created using NLP techniques.

  • Accuracy achieved using Bag of words technique on stemmed data is 98.56%.
  • Accuracy achieved using Bag of words technique on lemmatized data is 98.29%.
  • Accuracy achieved using TF-IDF technique on stemmed data is 97.93%.
  • Accuracy achieved using TF-IDF technique on lemmatized data is 97.93%.

  (the accuracy for tfidf model for stemmed & lemmatized data is same but confusion matrix was different)

• Created a pickle file for model with highest accuracy i.e. Bag of words technique on stemmed data (98.56%).

• Build Graphically user interactive front end web pages.

• Used Flask web framework to generate the web app using pickle file and front end web pages.

• Deployed the web app on Heroku platform.

• Enter the message in the empty box.
• Click on predict.
• Result will be displayed on your screen.