This project analyzes how cryptocurrency news headlines influence Bitcoin price movements. We collected real-time headlines using the CryptoPanic API and evaluated their sentiment with models like FinBERT and CryptoBERT. Using decision models such as Random Forest, we tested how well sentiment predicts price direction. Automated daily data collection was done using Google Apps Script. By varying sentiment and decision models, we aim to identify the best combination for accurate Bitcoin trend prediction.

This study examines how Elon Musk’s tweets influence short-term stock prices using AI-based sentiment analysis. We find that positive tweets often coincide with price increases, while negative tweets link to volatility. Crucially, similar messages from less influential people do not create the same effect, showing that sentiment alone cannot predict market behavior—the sender’s influence also matters.

This paper discusses the methods and results of our analysis of the factors in determining the winner of the popular online game Basketball Heads. We collected data on 28 players and used decision trees to determine the most important factors. Our analysis shows that ball possession is the most important factor, with an ~80% accuracy in determining player outcomes.

This project uses AI and computer vision to estimate fruit sweetness (brix index) from ordinary RGB and multispectral images. We built models that predict grape sugar content from photos and tested whether adding details like variety and harvest time improves accuracy. What makes this work promising is that it can function with simple smartphone images, allowing farmers to assess grape quality directly in the field. We tested five grape varieties—including Itum, Autumn Royal, and Crimson—to study how color and texture relate to sweetness, showing the potential for fast, low-cost fruit quality evaluation without lab equipment.

This paper introduces Homomorphic Broadcast Encryption (HBE), a unified framework that combines homomorphic encryption for privacy-preserving computation with broadcast encryption for selective data access. We implement HBE in cloud and IoT settings to enable secure aggregation and controlled result sharing. Tests on a 9-node cloud with 12 users show efficient encrypted computation, fast revocation, and low latency. HBE offers a practical solution for secure collaborative processing in areas like healthcare analytics and financial risk assessment.

This project develops an efficient method for automating water quality assessment using satellite images and a lightweight deep learning model. We use a CNN enhanced with Class Activation Maps (CAM) to detect water pollution while keeping the architecture small—three convolutional blocks, global average pooling, and a simple classifier. Trained on datasets of polluted and natural water bodies, the model achieves high accuracy with far fewer parameters than existing solutions, making it a practical and scalable tool for future pollution detection.