SMS_Spam_Detection_using_NLP

A supervised deep learning project using Natural Language Processing(NLP) that classifies a given SMS as a spam/ham message.

Dependencies/Libraries used:

• Pandas
• Numpy
• Seaborn
• Sklearn
• NLTK
• Pickle
• Streamlit

Input data:

Dataset - https://www.kaggle.com/datasets/bagavathypriya/spam-ham-dataset

Building Model:

• Imbalanced Data

• Visualizing the trend of number of characters, number of words and number of sentences in a spam/ ham SMS

• Plotting the trend as a Heatmap

• Most common words in a Spam Corpus

• Performance of Various Models

Results:

• Applied NLP techniques like tokenization, lemmatization, and stop words and punctuation removal using NLTK and regex.
• Performed feature engineering with handcrafted features such as digit count and email length.
• Implemented various classification models, Naive Bayes, SVC, ETC to find the best performer.
• Created an ensemble model, improving the accuracy from 97.87% to 98.25%.
• Designed & deployed a basic UI with Streamlit for classifying new inputs as spam or ham.

Top Performing Model

Combined Support Vector Classifier, Multinomial Naive Bayes and Extra Trees Classifier to build an Stacking Classifier Model( Ensemble Model) whose:

• Accuracy: 0.9825918762088974

• Precision: 0.9736842105263158

Testing Examples:

• Example 1:

• Example 2:

• Example 3: