The idea behind solving word problems is to develop a computational system that can understand, interpret, and solve mathematical problems presented in natural language. The word problem solving typically involves translating textual information into mathematical expressions and equations, a task that requires both linguistic and mathematical understanding. In our research, we have decided to focus on percentage-based mathematical problems. To solve these problems within a financial context, we have applied a rule-based approach using natural language processing (NLP). Our approach involves extracting relevant features and data using regular expressions, as well as recognizing synonyms to better understand diverse inputs. Overall, we aim to create a flexible and effective system for tackling word problems in this domain. The model is tested with our prepared dataset, which contains a variety of financial problems, including finding simple or compound interest, estimating time length, and calculating the principal amount, all of which are commonly used in the banking sector. The system outperforms other models, achieving an accuracy rate of 80% in solving financial math word problems.