InvestED: A Smart Financial Education Platform Using LSTM & Game Mechanics

In the era of state-of-the-art machine learning, Long Short-Term Memory (LSTM) networks and Large Language Models (LLMs) have shown remarkable capabilities in sequence prediction and natural language comprehension respectively. Although such technologies have remade several domains, their application in financial education remains unexplored. In this research, we present InvestED, a new platform that taps the temporal dependency modeling capabilities of LSTM neural networks and contextual knowledge of LLMs to redefine financial literacy education for students. The bidirectional property of LSTM enables sophisticated market patterns to be captured by its dedicated forget, input, and output gates, while transformerbased LLMs provide natural language processing for personalized financial suggestions. Our setup utilizes the double-layer LSTM model ($\mathbf{1 0 0}$ and $\mathbf{5 0}$ units) to predict the performance of mutual funds, beating conventional statistical frameworks such as ARIMA and Prophet by $\mathbf{2 9 - 4 4 \%}$ lower error rates for all mutual fund types. The platform includes such advanced AI models within game-like interfaces such that a computation arena is availed to test investment decisions, monitor expenses, and learn financial aspects through the feedback loops produced by algorithms. Performance evaluation demonstrates the neural network’s superior potential for identifying non-linear market dependencies with mean absolute percentage errors as low as $\mathbf{2. 3 1 \%}$ for certain classes of investments in out-of-sample tests. InvestED demonstrates the potential of advanced deep learning architectures and natural language processing to redefine abstract financial education as a fun, data-driven learning experience, effectively bridging the pedagogical and technological challenges of equipping digitally-native generations with financial literacy.