This project focuses on classifying SMS messages as Spam or Ham (Not Spam) using various machine learning models and text processing techniques in Python. The pipeline includes preprocessing, exploratory data analysis, vectorization, model training, evaluation, and model saving.

• pandas

  • read_csv, drop, rename, apply, value_counts, describe, isnull, duplicated, drop_duplicates

• numpy

  • For numerical operations (imported but not heavily used)

• matplotlib

  • pyplot, figure, show

• seaborn

  • histplot, pairplot, heatmap, catplot, barplot

• wordcloud

  • WordCloud class to visualize most common words

• nltk (Natural Language Toolkit)

  • word_tokenize, sent_tokenize, stopwords, PorterStemmer, download

• string

  • Used to remove punctuation

• Preprocessing

  • LabelEncoder for encoding target labels

• Data Splitting

  • train_test_split to split data into train and test

• Evaluation Metrics

  • accuracy_score, confusion_matrix, precision_score

• Models Used

  • 🧠 Naive Bayes: GaussianNB, MultinomialNB, BernoulliNB
  • 📈 Logistic Regression: LogisticRegression
  • 🔍 Support Vector Machine: SVC
  • 🌳 Tree-Based Models: DecisionTreeClassifier, RandomForestClassifier, ExtraTreesClassifier
  • 📊 Ensemble Methods: AdaBoostClassifier, BaggingClassifier, GradientBoostingClassifier
  • 👥 Others: KNeighborsClassifier, VotingClassifier, StackingClassifier

• Feature Extraction

  • CountVectorizer, TfidfVectorizer

• pickle
  • dump() to save model and vectorizer
  • load() to load them later for predictions

• collections

  • Counter class used for word frequency count

• warnings

  • Used for ignoring unnecessary warnings during model training

• Multiple models were trained and evaluated
• Accuracy and precision were used as evaluation metrics
• The best performing model was saved using pickle

Create a requirements.txt with: