Forecasting loan disbursement is essential for financial institutions to manage liquidity effectively and mitigate risks. Traditional forecasting methods often struggle with the complex and seasonal dynamics present in financial data. This study compares several forecasting models including ARIMA, Linear Regression, Long Short-Term Memory (LSTM) networks, and Prophet using seven years of loan disbursement data. Results indicate that LSTM networks produce the most accurate forecasting. The findings show the potential of machine learning to improve planning and decision-making for lenders, while maintaining strict considerations for data privacy and fairness.