Emotion Aware Recommender System
Recommendations are a rapidly increasing paradigm that automatically provides and assists the user in finding what he or she is seeking from a vast amount of data. A recommender system is the name for this paradigm. In many circumstances, traditional recommender systems have worked effectively; howe
2025-06-28 16:26:59 - Adil Khan
Emotion Aware Recommender System
Project Area of Specialization Artificial IntelligenceProject SummaryRecommendations are a rapidly increasing paradigm that automatically provides and assists the user in finding what he or she is seeking from a vast amount of data. A recommender system is the name for this paradigm. In many circumstances, traditional recommender systems have worked effectively; however, many recommender systems ignore important information such as location and time which are the contextual elements (information) in a recommender system. Recommender systems that do integrate contextual information are called context-aware recommender systems. A standard context-aware recommender system will have at most 2 contextual elements for the purpose of generating recommendations, and although they have been successful, they still lack a certain degree of accuracy and result generations. The main focus of this project is to incorporate into these context-aware recommender systems a third contextual element which is the emotion of the user. A user's current emotion is a very valuable piece of information which can be used for more accurate recommendations. Through this recommender system we will be able to capture the user's emotions either through a smartphone camera or prompt input and present them recommendation based on those current emotions and alongside the other context provided. This recommender system is aimed at the entertainment industry, to give users a platform for movie content recommendations through their emotions. The main project will be deployed onto a Mobile Application with a webpage for information on how the App is to be used. Furthermore, it will pave a new way to how recommender systems are utilized and how to improve them in big companies such as Netflix, Hulu, Amazon Prime etc. The market value of recommender systems has been high ever since they came out and any advancements in them have always produced favorable results and figures. The Emotion Aware Recommender System will be one of the first of it's kind.
Project ObjectivesThe objective of the Emotion Aware Recommender System are:
- To analyze existing data and performing pre-processing steps on them.
- To define algorithms for user profiling and emotion detection.
- To investigate the precedence of data vs emotional state.
- To build a Web Page and a Mobile Application Recommender System (Based on emotions).
The recommender system will be made user accesible through a mobile application. The system will be using the typical three main aspects : Users, Items and ratings but will now also include a new contextual element which is the user's emotion. The user's emotion will be captured using the smartphone camera and the image will be processed through a model to generate results which will then be used to calculate recommendations. The web page will serve as the information page for the Application and as a location to download the app itself.
Benefits of the ProjectThe beneficiaries of this project are:
- Users (Recommendations for movies based on their emotions)
- Entertainment Industry (Accurate User recommendations will generate more clicks and more reveneu.)
- Other stakeholders:
- Researchers
- Developers
Recommender Systems are based upon algorithms which can be categorized into three main branches:
Collaborative Filtering (CF): The most common application of collaborative filtering (CF) algorithms is to create a recommender system.
Large volumes of data about a user's behaviour, activities, and preferences are collected and analysed via Collaborative Filtering (CF) to better predict what a person likes based on their resemblance to other users.
Content Based Recommendations: The information in content-based recommendations comes straight from the contents of the items rather than the user's opinions. A machine learning technique is used to model a user's preference data from instances, based on the description of the contents, where the contents of things are the explicit features or characteristics, such as a movie's genre, feature, and release year.
Hybrid Recommender Systems: Hybrid recommender systems are more complex recommender systems that integrate two or more recommendations algorithms to improve performance.
The project's recommender system will be a blend of the above three categories.
The application will be built using React Native so as to be supported on both Android and iOS smartphones. Python will be used for creating the CNN models for emotion detection and recommender system algorithms. A mySQL database will be employed for storing information about the user and their recommendation profiles. An API (Application Progamming Interface) will be built for proper communication between the React Native Frontend and the Python+PHP backend.
The web page will be a simple single page form which will showcase the information about the mobile app and how it works. It will also have the option to download the mobile application.
| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 10000 | |||
| Google Colab Pro Plan | Miscellaneous | 1 | 6000 | 6000 |
| Overhead (Currency Rate Issues, Stationery) | Miscellaneous | 1 | 4000 | 4000 |