This study examines the time series forecast algorithms for fixed rate mortgage average in the United States using weekly data from 2019 to 2024 and compare their performances via Mean Absolute Percentage Error (MAPE). Five models—ARIMA, ETS, TBATS, NNAR, and Prophet —were trained and assessed. The findings revealed that the NNAR model outperformed others with the lowest MAPE (6.19%) and proved its capability to capture non-linear trends. ETS also showed admirable performance with better forecasts from the remaining models. Despite normality assumption violations, most models achieve lack of serial correlation and constant variance assumption.