Deep Learning Models for Solving Financial Word Problems

Deep Learning Models for Solving Financial Word Problems

연구 분야: Artificial Intelligence

학회: SN Computer Science

The integration of artificial intelligence across different fields has recently brought about significant transformations, with deep learning models playing a crucial role in addressing complex mathematical challenges. This paper introduces a new dataset to help deep learning models for solving financial word problems (FWP). The dataset includes common financial concepts, such as calculating simple and compound interest, estimating time duration for loans or investments, and determining principal amounts. These types of problems are highly relevant to banking and finance. The dataset is organized in a structured way, making it easier to extract key information for training deep learning models. We conducted a comparative study of different deep-learning approaches. Various models were tested to see how well they could solve the financial word problems. Among the models, we found that Long Short-Term Memory (LSTM) networks performed the best on our dataset. This result suggests that LSTMs are particularly well-suited for financial problem-solving tasks where understanding sequential and contextual information is crucial. In addition to developing the dataset, this research highlights the potential for deep learning in the financial domain. By providing a specialized resource for financial word problems, we aim to support further research and development in AI-driven financial solutions. This work not only contributes a new tool for training models but also demonstrates the practical applications of AI in solving complex financial scenarios.