This repository is based on the flask framework web application, with the help of Machine Learning, the application is capable to detect that the comment is spam or not.The goal of our project was to create a machine learning model that can produce output for anyone. Our project has an integrated API structure that recieves user's queries and responds to them with the output of our spam comment classifier model when fed with the query input parameter.
http://spamcommentdetector.herokuapp.com/predict?commentInput=parameter
It will return a json object having input and outcome key which will have input given and resultant outcome eithier of the value Spam or Not a Spam
📦Spam-comment-detector-with-flask
┣ 📂static
┃ ┗ 📂js
┃ ┣ 📜app.js
┃ ┣ 📜form.js
┃ ┗ 📜particles.js
┣ 📂templates
┃ ┗ 📜form.html
┣ 📜SPAM.csv
┣ 📜app.py
┗ 📜requirements.txt
A microframework based on Werkzeug. It's extensively documented and follows best practice patterns.
Convert a collection of text documents to a matrix of token counts.This implementation produces a sparse representation of the counts using scipy.sparse.coo_matrix.
If you do not provide an a-priori dictionary and you do not use an analyzer that does some kind of feature selection then the number of features will be equal to the vocabulary size found by analyzing the data.
Split arrays or matrices into random train and test subsets. Quick utility that wraps input validation and next(ShuffleSplit().split(X, y)) and application to input data into a single call for splitting (and optionally subsampling) data in a oneliner.
The multinomial Naive Bayes classifier is suitable for classification with discrete features (e.g., word counts for text classification). The multinomial distribution normally requires integer feature counts. However, in practice, fractional counts such as tf-idf may also work.
-
Click==7.0
-
Flask==1.1.2
-
pandas==1.1.0
-
sklearn==0.0
-
joblib==0.16.0
-
gunicorn==20.0.4
-
itsdangerous==1.1.0
-
Jinja2==2.10.3
-
MarkupSafe==1.1.1
-
Werkzeug==0.16.0
