Scribd
Upload
11 views4 pages

Cinelens Ai Project

The paper presents a novel movie recommendation system that utilizes user behavior analytics and advanced machine learning techniques, specifically matrix factorization and KMeans clustering, to provide personalized suggestions. A proprietary browser extension extracts relevant keywords from users' browsing history, which are then processed by a hybrid ML model to enhance recommendation accuracy, achieving a training loss of 0.0025 and an accuracy of 90.12%. The system aims for real-time recommendations and plans to expand its capabilities across more browsers to improve user experience and scalability.

Uploaded by

Sayandeep Das
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
11 views4 pages

Cinelens Ai Project

The paper presents a novel movie recommendation system that utilizes user behavior analytics and advanced machine learning techniques, specifically matrix factorization and KMeans clustering, to provide personalized suggestions. A proprietary browser extension extracts relevant keywords from users' browsing history, which are then processed by a hybrid ML model to enhance recommendation accuracy, achieving a training loss of 0.0025 and an accuracy of 90.12%. The system aims for real-time recommendations and plans to expand its capabilities across more browsers to improve user experience and scalability.

Uploaded by

Sayandeep Das
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

CineLens: Advanced Movie

Recommendations Using Matrix Factorization


& KMeans Clustering

Team Members: Rohit kumar singh(22BRS1258) Sayandeep das (22BRS1044)

Abstract
In this paper, we present a new end-to-end solution for receiving personalized movie
recommendations leveraging user behavior analytics and advanced ML techniques. Using a
proprietary browser extension, our system scans users' browsing history to automatically
extract useful keywords indicative of genre and emotion. For working, further the keywords
are fed into a machine learning model specially built using matrix factorisation and KMeans
clustering to suggest movies closely related to the user interest. At the heart of our system is
the hybrid ML model, which incorporates some matrix factorization on user ratings to make
recommendations as well the KMeans clustering (of five clusters) for movie
recommendations based upon keyword similarities. By combining behavioral insights with
more structured information on user preferences, this method allows for improving
recommendations. After many possible training and hyperparameter testing, Our model
achieved a training loss of 0.0025, test loss of 0.0031 and an accuracy of 90.12% which was
greater when compared to traditional recommendation methods. For browsing ease and
optimal extraction, the system shall include a user-friendly browser extension for seamless
keyword extraction, while the model itself is highly efficient which allows it to provide
recommendations in real-time. Long term improvements will include the ability to provide
near real-time updates and broader availability across more browsers in order to increase both
the accuracy of results and scope of scalability. Such integrated framework is effective in
capturing user preferences and enhancing overall user experience by providing personalized
movie recommendations based on historical data of individual users.

Keywords- Movie Recommendation, Machine Learning, Matrix Factorization, KMeans


Clustering, User Behavior Analytics, Browser Extension, Personalization
1. INTRODUCTION
The increasing availability of digital content platforms and the growing popularity of
streaming services have led to a surge in demand for advanced recommendation systems1.
They are invaluable for user and designer experience by personalizing the experience to make
it engaging and leveraging maximum potential. Traditional recommendation models often
use user ratings and collaborative filtering methods. While these techniques can be effective,
they are inherently limited, as computation only takes into account part of user likability. In
order to deal with these constraints, a new recommender system on movies will be presented
in this paper by integrating behavioural analytics from users browsing data with sophisticated
machine learning (ML) and hence providing more advanced suggestions.
Our system uses a patented browser extension which helps to identify the information of the
users browsing history, from which we get relevant genre based keywords, emotional tone
and other attributes that are key indicators of interest. The keywords are passed into hybrid
ML model, Fox, which is comprised of matrix factorization for rating based filtering and
KMeans algorithms on keyword based co-occurrence clusters. For example, matrix
factorization is scalable and great at revealing underlying factors related to user—you can
use this for movies by looking at user ratings to predict movie preference. At the same time,
KMeans with five clusters is applied to form groups of movies based on thematic or stylistic
similarities at the level of singular keywords. Through this combination, it allows the system
to integrate static data (like genre or popularity) with real-time behavioral data and improves
its recommendation engine based on the users preference.
After all the proper hyper parameter tuning and optimization, this hybrid model was able to
produce a highly competent system with training loss: 0.0025 Test Loss: 0.0031 Accuracy :
90.12% Such accuracy is much higher than traditional recommendation methods, which
indicates that the system can generate reliable and personalized recommendations.Moreover,
the browser extension allows for quick and live keywords extractions that improve the entire
user experience by providing seamless continuous updates of recommendations. The model
architecture is designed to be much faster and more efficient, resulting in real-time
recommendations without substantial computation time.
We will be expanding the system to support additional browsers, as well as improving its near
real-time capabilities to make it more adaptable to changing user trends in future releases.
The system has improved accuracy and scalability by incorporating these features to ensure
that recommendations are accurate and relevant in accordance with evolving interests.
Behavioralbased integration with machine learning enhances user engagement and elevates
personalized recommendation framework standards leading to a friendlier digital experience.

2. KEY TECHNOLOGIES USED


The three key technologies which the movie recommendation system employs to achieve
personalization and relevance are:
1) Matrix Factorization (MF):
Matrix Factorization which falls under collaborative filtering studies the user-movie
interactions and factorizes a massive user-movie matrix into latent factors of users and
movies. The latent factors reflect user tastes and movie characteristics, enabling the
system to impute values and predict ratings for movies not yet rated. Due to its capability
to deal with sparse data, MF is capable of making personalized recommendations that
match the users and movie haters in an effective manner.

2) KMeans Clustering:
The movies are clustered into five clusters based on the thematic and keyword similarities
using KMeans clustering, an unsupervised learning algorithm. Every cluster has
representation of group of movies which could be a genre, mood-based or any theme-
based. The system suggests movies that correspond to a user’s cluster of interests, which
means recommendations are being targeted by the domain of interest. Through this
clustering, the recommendations will continue to be relevant as they are limited to the
particular movie groups that correspond with user preferences.

3) Keyword Extraction through Browser Plugin:


The user browses the web normally and our custom JavaScript browser extension keeps
track of their browsing to identify keywords relevant in relation to searches conducted.
Common keywords such as genre, mood or actor are captured and used to construct an
upto-the-minute profile of the user interests which allows for dynamic recommendations.
This boosts user privacy as data is processed locally inside the browser itself so sensitive
information remains safe.
The fusion of MF predictive power, contextual grouping with KMeans clustering and real-
time keyword extraction creates an adaptive recommendation system. Not only does this
hybrid model personalize the movie selection but it also adapts to any alterations in user
preferences making for a very streamlined and dynamic recommendation experience.

6.CONCLUSION
This research paper introduces a new method of improving movie recommendation systems
by merging user behavior analytics and state-of-the-art machine learning methods. We
propose a novel system that combines matrix factorization combined with clustering
algorithms as a browser extension to gather keywords from user browsing logs and
personalized recommendations based on user/psalm interest. Together, this technique
provides a more precise model fitted to the user so that the user obtains better quality
recommendations.
Our model demonstrates a promising approach to enhance the relevance of movie
recommendations by combining up-to-date observation data with the latent factor model,
achieving improved performance in terms of both relevance and user engagement. Instant
update and adaption of the enhancement today would cost in scalability drive before this
extension to other browsers even better its accuracy and scope. This system represents a

significant step forward in the ongoing evolution of personalized movie recommendations


and we look forward to even better methods for responsive, effective recommendation
engines that stand the test of use in real-world scenarios.

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy