Motive and Goal is developing a spam email classifier so that it can classify the spam email.
Data Collection
Data collected from: https://www.kaggle.com/datasets/venky73/spam-mails-dataset
This dataset contains huge data, almost 5171 Email found here.
Data Cleaning and preprocessing
This dataset contains lots of duplicate Emails. So, I drop the duplicate Email and in the preprocessing stage I remove the numerical symbol, number etc. This process reduces training time, facilitates
internal computation, and increases Accuracy.
FrameWork
For this project I use Google Colab, NLTK, gensim libraries. These are really helpful to implement code.
Model selection
In this project I've experiment with different models. Finally, SVC model provides, that gave me better result than another model.
I've tried KNeighborsClassifier, DecisionTreeClassifier, RandomForestClassifier and lastly SVC.
Models Comparison & Evaluation
| Model | Training Accuracy | Testing Accuracy | F1-Score | AUC Score | CV Score |
|---|---|---|---|---|---|
| KNeighborsClassifier | 83.65% | 81.67% | 72.60 | 82.28 | 0.7999 |
| DecisionTreeClassifier | 100% | 77.88% | 61.46 | 72.83 | 0.7864 |
| RandomForestClassifier | 100% | 88.40% | 76.85 | 81.89 | 0.8788 |
| SVC | 98.25% | 94.28% | 89.97 | 92.58 | 0.9241 |
As I want to make an app, I have to deploy my project. So I've chosen the Streamlit Cloud to deploy.
You can find my work in here: https://alaminbhuyan-email-classification-main-oaiqrp.streamlit.app/
| Index | Comment |
|---|---|
| 1. | Recommended IDE: PyCharm Community edition |