/TheSystem_SpamDetection

Android app for spam and fake review detection.

Primary LanguageKotlinMIT LicenseMIT

The System - Android Application

Team, The Developers's project for Codathon event for the 6th CSI JK STATE STUDENT CONVENTION 2020.
Update: This Project was awarded the FIRST prize

Features:

  1. Login as Admin or User.
  2. OTP Authentication
  3. Add new product (only for admin). Admin can choose Name, description, price and Image for product
  4. View products
  5. Write a review (only for user).
  6. Delete review, block/unblock user (only for admin)
  7. Notification, if new spam review is detected (only for admin)
  8. Remove Products

Download APK

Note: This app is not fully tested so unexpected behaviour may occur in some devices.

Spam detection

A sophisticated system is built to detect spam reviews. Detection process involves two phases
  1. Logically detecting if user has reviewed multiple times on the same product
  2. Machine learning classification model (using Tensorflow lite) is used to analyze text and classify review into truthful or deceptive. Dataset is downloaded from Kaggle
If a review is detected as spam in both the phases, then review is confirmed to be spam and Alert Notification is show to admin. If review is detected as spam in anyone of the phase, then review is marked as potential spam and admin is notified.

Check out Jupyter Notebook

Tensorflow lite is used to embed model into Android app.
Model resides in app/src/main/assets/model.tflite

Phone numbers for testing:

Phone number Verification code
+91 66666 66666 666666
+91 11 1111 1111 111111
+91 88888 88888 888888

Screenshots

Welcome Page Login via Phone
Submit OTP Submit details Page
Home page (Notification for admin) Add Product (admin feat.)
Write Review (user) Review Details (admin)
Navigation Reviews Page
When admin blocks the user and next time user opens app, gets this msg

Note: This app was developed in hurry because of limited time in hackathon. So bugs and issues are expected.