Modelling and forecasting of interest rates has traditionally proceeded in the framework of linear stationary models such as ARMA and VAR, but only with moderate success. We examine here four models which account for several specific features of real world asset prices such as non-stationarity and non-linearity. Our four candidate models are based respectively on wavelet analysis, mixed spectrum analysis, non-linear ARMA models with Fourier coefficients, and the Kalman filter. These models are applied to weekly data on interest rates in India, and their forecasting performance is evaluated vis--vis three GARCH models (GARCH (1,1), GARCH-M (1,1) and EGARCH (1,1)) as well as the random walk model. The Kalman filter model emerges at the top, with wavelet and mixed spectrum models also showing considerable promise.