AI-Powered Personal Finance Manager using Time- Series Forecasting and Anomaly Detection

Managing money has become harder because people are spending money in new ways, and there is a greater chance of fraud. Traditional budgeting tools often have limited ability to make predictions and analyse data, and the decision-making processes that go along with them may not be clear. This paper analyses a web-based personal finance manager that uses artificial intelligence to help people keep track of and manage their money. The goal is reached by using time-series forecasting, anomaly detection, and explainable artificial intelligence (XAI) methods all at once. Using historical transaction data, we use Long Short- Term Memory (LSTM) networks to predict future costs. Autoencoder models are often used to find potentially fraudulent or unusual activities in financial transactions. Pandas and NumPy are used to clean the data and make new features. After that, the models are deployed as RESTful services using a FastAPI backend. To help users trust the results and understand them better, SHAP and LIME methods are used to explain how the models make predictions and find unusual spending. Experimental results demonstrate the precision of expense forecasting and the detection of irregular spending patterns. The suggested system provides a scalable, safe, and easy-to-use way to help people make smart decisions about their money.