This project develops advanced models for stock allocation to maximize returns using both short-term and long-term portfolio optimization strategies. For short-term optimization, we combine sentiment analysis via the BERT language model with the Black-Litterman model for dynamic 10-day portfolio adjustments. For long-term optimization, we utilize a Long Short-Term Memory (LSTM) network to predict stock performance and compare it against the Markowitz Mean-Variance Model and Genetic Algorithm. These approaches aim to create a versatile toolkit for optimizing portfolios under varying market conditions and investment horizons.