The use of an artificial neural network (ANN) is becoming very common nowadays due to its ability to analyze complex nonlinear events. An ANN has a flexible, convenient and easy mathematical structure to identify the nonlinear relationships between input and output data sets. This capability could efficiently be employed for the different hydrological models such as rainfall-runoff models, which are inherently nonlinear in nature and therefore, representing their physical characteristics is challenging. In this paper, the influences of back propagation algorithm and their efficiencies which affect the input dimensions of the rainfall-runoff model have been demonstrated. The capability of the Artificial Neural Network with different input dimensions has been attempted and demonstrated with a case study on Sarada River Basin. The ANN models developed were able to map the relationship between input and output data sets used. The model developed on rainfall and runoff pattern have been calibrated and validated. The significant input variables for the training of ANN models were selected based on statistical parameters like cross-correlation, autocorrelation, and partial autocorrelation function. It was found that those models considering rainfall lag rainfall and discharge as inputs were performing better than those considering rainfall alone. It was found that the neural network model developed was performing well. It can be inferred from the developed model that the Neural Network model was able to predict runoff from rainfall data fairly well for a small semi-arid catchment area considered in the present study.