You are on page 1of 9

Major Project On

Content Based Advertisement


Recommendation System
2022-2023

Ashwani Kumar
Abhishek Chauhan
Vatsalya Baluni
Akrat Chandola
Presentation Flow

Algorithms
Used

Work Flow Applications

Features Future
Of Works
Projects

Project Conclusion
Objective
Project Objective

Introduction Why Advertising? Goals


We are constantly bombarded with data in the • Advertising is the only medium you can control
• To create a advertisement system which gives accurate

• Advertising allows you to target ideal customers only output based on user input.
modern digital era, including emails, social media
• Advertising creates awareness for your content
• Fast Recommendations.

feeds, news stories, and items to buy. Users • Advertising adds credibility to your message

consequently frequently struggle to locate • Advertising amplifies everything else you’re doing

pertinent and amusing information and goods that


match their tastes and interests. Systems that
make recommendations are useful in this
situation.
Computer programs known as recommendation
systems utilize machine learning algorithms to
analyze user data, including reviews, purchases,
and browsing history, to spot trends and generate
customized recommendations. These suggestions
can be for everything from items to potential
romantic partners to films, music, and literature.
Features
Of
Projects
Work Flow
Flow of Algorithm Flow of UI Pre-Processing
Flowchart
Algorithms Used
• TF-IDF with Cosine similarity
• TF-IDF with Pearson similarity
• TF-IDF with Euclidean similarity
• Bag a Words with Cosine similarity
• Bag a Words with Pearson similarity
• Bag a Words with Euclidian similarit
Applications
• E-Commerce and Retail
• Media and Entertainment
• Social Media
• Travel and Hospitality
• Health and Fitness
• Job Search
• Online Advertising
• Education
• Financial Services
• Food and Beverages
• Personalisation in Online News
• Online Dating
• Real Estate
• Automotive Industry
• Beauty and Cosmetics
Future Works
• Using deep learning to understand patterns in data for more personalized recommendations
• Incorporating contextual data such as location, time of day, and weather conditions into recommendation
systems can provide more relevant and personalized recommendations, enhancing the user experience.
• Improving collaborative filtering, which is how recommendations are made based on what other people with
similar interests have liked. Overcoming the limitations of collaborative filtering, such as the cold-start problem,
can result in more precise recommendations based on user behaviour and preferences.
• Future research can focus on developing more transparent and explainable recommendation systems to establish
user trust and enhance user satisfaction. As systems become more sophisticated, users want to know why a
particular recommendation was made.
• Using recommendation systems in different fields, like healthcare or education, to suggest personalized
treatments or learning materials.
Conclusion
In this project we tried to create a content-based recommendation system in which we applied various mathematics techniques such as vectors-based model and bag
overs to achieve it. The project we have created using term frequency and inverted term frequency. The goal of our project was to create a content-based
recommendation system in which we applied various statistical techniques such as vector space model and similarity techniques.

In this research we have compared the proposed method with various other methods to find the difference between the different methodology. The dataset used in this
research is flipkart products dataset which consist of thousands of products. The dataset is first pre-processed for tags and the different vector-space representation
methods are used with variation of different similarity algorithms.

On comparing the similarity matrix of different product tags, we created a hotspot map of two vector-space representation methods that are bag of words and TF-IDF.

We can compare the different methodologies using the correlation analysis. Correlation coefficient will indicate the degree of similarity between the two similarity.

Methodologies Correlation constant with the proposed method


TF-IDF with Euclidean similarity 0.8280
TF-IDF with Pearson similarity 0.9570
Bag a Words with Cosine similarity 0.8044
Bag Of Words with Euclidean similarity 0.4852
Bag Of Words with Pearson similarity 0.7526

Table 1 Corelation comparison between different Methodologies

You might also like