Geomagnetically induced currents (GICs) are low-frequency quasi direct currents caused by the complex interplay between Earth geomagnetic field and continuous high speed solar wind stream (HSS). The result of this interaction poses significant threats to the operations of technological infrastructures such as power grids. The integrity of oil pipelines may also be affected by the influx of GICs. As such, developing models for accurate and timely forecast of GICs is essential for mitigating the potential hazards and safeguarding valuable systems from the extreme effects of these currents. In this work, we propose a machine learning model based on the Nonlinear Autoregression with Exogenous inputs (NARX) for forecasting GICs. The model developed here forecasts GICs using inputs solely based on the solar and interplanetary parameters. The solar and interplanetary parameters retrieved from Omni Web during the maximum and minimum phases of solar cycle 23 were utilized. The developed model takes the solar wind speed, Bz, IMF, AE, ASYH, and SYMH as inputs and the measured GICs obtained from the Finnish Meteorological Institute (Mäntsälä) as the target. We validated the model using measured GICs during the geomagnetic storm periods of November 20 -23 2003, November 7 -13 2004 and August 24 -26, 2005. The model's performance was evaluated using the cross-correlation coefficient (R), root-mean square error (RMSE)and the wavelet coherence analysis. The prediction accuracy for each of the individual storms are 69 %, 68 %, and 70 %, respectively. For these events, RMSEs of 1.17 A, 1.58 A and 0.56 A respectively were obtained in each case indicating the robustness of the model. The approach presented augments existing works and will contribute to the forecasting of GICs in areas where the geoelectric field and local geophysical parameters are not readily available.