Introduction

Conexus is a social media application that aims to foster genuine connections by bringing together people with shared interests through personalized community recommendations. Unlike conventional social media platforms that focus on metrics like followers and likes, Conexus is designed to promote real-world interactions and meaningful engagement. By utilizing machine learning algorithms and natural language processing models, the app helps users discover communities, events, and individuals that align with their personal and professional goals.

Objectives

1. Develop a Community-Driven Social Platform
   Build an application that connects individuals based on shared interests, focusing on multicultural communities in Hong Kong.
2. Implement a Personalized Matching Algorithm
   Create an intelligent recommendation system that uses machine learning to identify and suggest communities based on user profiles.
3. Build a Predictive Event Recommendation System
   Develop a machine learning model that suggests events tailored to users’ preferences, encouraging real-world interactions.
4. Ensure User Safety through Real-Time Content Moderation
   Use natural language processing techniques to implement real-time spam and abuse detection.
5. Deliver a Cross-Platform Mobile Application
   Use the Flutter SDK to support both Android and iOS devices, ensuring a consistent user experience.

Methodology

The project uses a combination of machine learning, mobile development frameworks, and backend cloud services to deliver a seamless, real-time social networking experience. Each key feature is developed with a focus on ease of use, security, and scalability.

Development Frameworks and Tools Used:

• Frontend: Flutter SDK for cross-platform app development on Android and iOS.
• Backend: Firebase for database, authentication, and real-time messaging.
• Machine Learning Models: TensorFlow Lite for on-device models, Matching Algorithms, LLMs, NLP models

Technical Process Flow:

1. User Registration and Profile Creation
   • Users create profiles and select interest tags.
2. Community Matching
   • The app recommends communities using clustering algorithms based on shared interest vectors.
3. Event Recommendations
   • LLMs suggest events tailored to user preferences.
4. Spam and Abuse Detection
   • Text and media exchanged in real-time are moderated using NLP models.