Other Projects
Interactive Map for Finnish Statistics
Overview
This project encompasses the development of a web-based platform designed to offer users a dynamic view of Finnish statistics through an interactive map. The map, built with Leaflet and enhanced with GeoJSON for city boundaries, allows users to explore a wide array of statistics for each Finnish city, such as population figures, marriages, and more. Featuring intuitive pop-ups, users can delve into detailed data and even export statistics of interest as SVG files. The application, responsive across desktop and mobile devices, integrates data from the stat.fi API, ensuring up-to-date and reliable statistical information. (Pictures coming soon)
Spam Email Detection with DistilBert
Overview
In our project, we employed the DistilBert model, a compact yet powerful derivative of the BERT architecture, to develop an efficient spam email detection system. We focused on optimizing DistilBert for high accuracy and computational efficiency by fine-tuning it with a specialized dataset of labeled emails. Our objective was to create a model capable of accurately differentiating between spam and legitimate emails, utilizing the advanced NLP capabilities of transformers without the need for extensive computational resources. This endeavor aimed at pushing the boundaries of spam detection technology, balancing precision with practical applicability.
Results
The fine-tuned DistilBert model demonstrated outstanding accuracy in spam email classification, achieving a 99.32% success rate. While it significantly outperformed conventional machine learning methods with a precision of 99.03% and recall of 98.71%, it's important to note the potential limitations in generalizability. These remarkable results were obtained on a specific dataset, and performance may vary on different email datasets. This caveat underlines the need for further testing and adaptation to ensure the model's effectiveness across diverse spam detection scenarios. (Pictures coming soon)
