System Introduction

This software is Based on Recommendation Engine, in which we are building a news recommendation system and Fake news detection software. The purpose of this software is to show or provide only that news to our users which is according to their interest and make sure that only authenticated news will be shown to our users

Background of the System

If we look at existing systems there are multiple platforms and news channels that give a lot of news but there are giving every news to every user, yes they have classifies the different categories of news and place and show in managed format but still, the user needs to find out the news of his/her interest suppose a person like to see a sports news so he/she needs to go in the sports section to see sports news. So, in our case, we are trying to reduce this searching problem and show only that news to the user which are related to the interest of that user.

If we look at social media there are a lot of platforms which are providing numerous amount of news to users but in some cases, they are also passing fake news (maybe for increasing the network traffic) to their user so we will be overcome this problem and make sure no fake news is passed to our users.

Objectives of the System

• User Don’t need to search for news, which is according to his interest, we will provide news according to their interest.
• Fake news should be avoided to give a user.
• Ringing notifications to the user's timetable.
• Provide Pakistan and world news on a single platform as user-friendly as it can.

Significance of the System

Today due to social media and easy access over the world there is a huge amount of news in the different category are available on different platforms on daily bases so due to the huge amount and category it is difficult for the user to stay updated on that news which is related to his/her interest or field or department, we are solving this problem by using a recommendation system of news, as we see there are multiple recommendation systems on other products but news recommenders are not often seen in the market in the so, now we are working on news recommendation system so our user is easily stay updated about what is now happening in Pakistan and world in their fields. As news is text data so our system can easily adapt to another field that has text data and our system will be easily integrated also with another domain. Due to the freehand on social media, it is very to spared fake news over the whole globe in seconds, so we also integrating a fake news detection module in our system so we can prevent our user from seeing fake news.

Over all working of the System

Initially I wrote web scrapers using python (scrapy), for each news provider in Pakistan e.g. Ary | Geo | Dunya | Bol | Daily | AbbTak etc. so firstly I was only scraping classified News (labeled News) and dump into our db (MongoDB Atlas) this process goes on for 4-5 months and we got 100,000 news of 15 different categories e.g Pakistan | World | Bussiness | Entertainment | Technology etc. Then I apply Data Wrangling concepts and clean data then by using NLP concepts and Random Forest I build model for news classification its accuracy was 74% during building model I faced Imbalnced classification problems. After building model I also started scraping the non labeled news and classifies them using our model this process goes on I used Apache AirFlow to build this piple line for scraping to classification and dumping into our centralized data base which runs after 3 hours every day.

Then for News recommendation I build dumpy web pages and ask peoples to select any 5 to 7 categories of news they want to see so for This purpose we ask peoples in our circle due to this activity we got 1,000 responses, based on those responses I used Apriori Algorithm for rules genration. Then we build WEB and Android application, Android application was build my patner and web app was build by me I used MERN stack for building web app. So in web app we ask user to signin and tell us any 3 categories they like based on that we build users profile using those rules we genrated and infomration given while singin then we use that infomration to recommend news, if user does not want to sign in we initialy show latest news according to their location and then we track their activities on our App and we use that infomration to build their profiles using local storage and cookies this mechanism is used by both Mobile app as well as Web app.

Click here to see Complete Report

• Undertaken By : Usman Ghani Mughal & Ume Zara
• Supervised By : Dr. Hikmat Ullah Khan
• University : Comsats University Islamabad
• Year : 2020-2021

Images WEB App

Images Mobile App