Spam SMS Classification 🔔 ⚠ ✔

Libraries

1. NumPy
2. Pandas
3. Matplotlib
4. Scikit Learn

• Feature Extraction : Count Vectorizer and TF IDF Vectorizer
• Model Selection : Train Test Split, Grid Search Cross Validation and K Fold Cross Validation
• Ensembles : Random Forest Classifier and Gradient Boosting Classifier

5. NLTK ( Corpus, Stopwords, Porter Stemmer, Word Net Lemmatizer )
6. String ( Punctuation )
7. RE : Regular Expression

Features

1. SMS ( Messages Extracted in form of TSV )
2. Labels ( Spam or Ham )

Machine Learning Pipeline

1. Import Data ( SMS )

2. Exploratory Data Analysis

3. Data Cleaning

• Remove Punctuation.
• Change to Lowercase.
• Tokenization : Splitting a Phrase | Sentence into List of Individual Words called Tokens.
• Remove Stopwords : Most Common Words ( Filtered out before processing Natural Language Data )
• Stemming ( Speed ) | Lemmatization ( Accuracy ) : Reduce the Word to its Base | Stem Form.

4. Feature Engineering

• Vectorization : Convert Text to Numbers ( Feature Vectors ) that an Algorithm and a ML Model can Understand and Learn.
• CountVectorizer : Extract Features from Text ( Count Occurence of Text in Corpus and Consider each as Feature | Column )
• Bag of Words : Count Occurence of the Word in each Document and each Word becomes Feature Represented by a Vector.
• TF-IDF : Represents Importance of the Word in the Document.
• Term Frequency : Number of Time the Term Appears in the Document.
• Inverse Document Frequency : Number of Documents containing the Word.

5. Fit | Train Base Model

6. Tune Hyperparameters and Evaluate with Grid Search Cross Validation

7. Final Model Selection ( Comparing Models and Selecting Best One ) Accuracy : 97%

Process

1. Split the Data into Training Set and Test Set.
2. Train Vectorizers on Training Set and Use that to Transform Test Set.
3. Fit Best Random Forest Model and Best Gradient Boosting Model on Training Set and Predict on Test Set.
4. Evaluate Results of these Two Models to Select Best Model.