A Natural Language Processing with SMS Data to predict whether the SMS is Spam/Ham with various ML Algorithms like multinomial-naive-bayes,logistic regression,svm,decision trees to compare accuracy and using various data cleaning and processing techniques like PorterStemmer,CountVectorizer,TFIDF Vetorizer,WordnetLemmatizer. It is implemented using LSTM and Word Embeddings to gain accuracy of 97.84%.
| Text Preprocessing Type | Logistic Regression | Multinomial NB | Support Vector Machine | Decision Tree |
|---|---|---|---|---|
| TFIDF Vectorizer + PorterStemmer | 96.68% | 97.30% | 98.47% | 96.68% |
| CountVectorizer + PorterStemmer | 98.65% | 98.56% | 98.74% | 97.84% |
| CountVectorizer + WordnetLemmatizer | 98.56% | 98.29% | 98.38% | 97.75% |
| TFIDF Vectorizer + WordnetLemmatizer | 96.41% | 97.48% | 98.47% | 96.86% |
- The dataset used is SMS Spam Dataset created by UCI Machine Learning.This dataset is downloaded in kaggle.You can download it here.
- Reference for this dataset can be found here
For any kind of suggesstions/ help in models code Please mail me at ksdkamesh99@gmail.com.