This project leverages Deep Learning and satellite imagery to develop an automated system for predicting and mapping poverty levels across different regions of Nigeria, providing a cost-effective alternative to traditional household surveys using CNN and transfer learning.

This project replicates and extends Acemoglu et al. (2008) on the relationship between income and democracy using advanced panel data econometric techniques including Fixed Effects, IV, and GMM estimation. Key finding: the income-democracy relationship disappears once country-specific factors are controlled.

This project develops and compares RNN and LSTM architectures for stock price prediction, with LSTM significantly outperforming RNN in accuracy (37% improvement in MAE, 94% R² score). Features an interactive Streamlit web application for real-time forecasting with customizable parameters, visual analytics, and CSV export capabilities.

The project focuses on developing an advanced deep learning model utilizing Long Short-Term Memory (LSTM) architecture for the recognition and classification of traditional Mauritanian music notes. Traditional Mauritanian music, deeply rooted in history and cultural heritage, features the tidinit instrument, renowned for its intricate melodies and rhythms. Central to Mauritanian music are the “maqamat,” which are musical modes or scales that organize melodies according to established rules and dimensions, imparting distinct characteristics and emotions to each.

This project focuses on the development of an advanced natural language processing (NLP) model utilizing Long Short-Term Memory (LSTM) architecture for French to English translation. Translation between these languages presents unique challenges due to differences in grammar, syntax, and vocabulary. By harnessing the power of deep learning and LSTM networks, the project aims to automate and enhance the accuracy of translation processes, catering to diverse applications ranging from literature and academia to business and communication.

Problem: The main challenge is to develop an accurate and robust prediction model capable of classifying users based on their propensity to convert on the site. This classification is crucial for guiding marketing initiatives and user experience enhancements aimed at maximizing conversion rates and, by extension, site-generated revenues. Proposed Solution: This data science project aims to develop a conversion rate prediction model using advanced techniques such as logistic regression. By following a rigorous methodology including data preparation, selection of relevant variables, model construction, and performance evaluation, our goal is to provide the marketing team with precise predictive tools to optimize campaigns and conversion strategies.