Currency crises in developing countries have been an attracting point for economists for the last couple of decades. Different analyses and me-thods have been employed to capture the timing and the effects of crises on foreign exchange markets. In this paper, we propose a new approach to the analysis. This approach comprises: a) Basic exploratory analysis of macro-economic time series data, b) Rule Ba-sed Fuzzy System Modelling (RBFSM) to capture the underlying behavior of selected macroeconomic variables. The predictive power of this methodology is tested on the Turkish data. Exploratory data analysis includes some statistical measures, Fourier analysis and Long Range Memory analysis of selected (available) macroeconomic time series data. After determining variables that deem to be important for large currency swings, data are split into two groups for training and validation of the model. Training data is clustered by Fuzzy System Modeling, FSM, schema which first applies Fuzzy C-Means (FCM) clustering algorithm to input and output space with several levels of fuzziness, m, and number of cluster, c, pairs. Amongst all m and c pairs the one that gives minimum root mean square error (RMSE) is selected. For the inference local OLS type Regression is used. This modelling can also be classified as a fuzzy data mining that uses a basic Fuzzy Rule Based System Modelling structure. FSM results are compared with GARCH/ARMAX and ANFIS model predictions.