Data quality in neural network models: Effect of error rate and magnitude of error on predictive accuracy

Neural networks have been applied in a wide variety of business domains. Although databases used in many organizations have been found to contain errors, little is known about the effect of these errors on predictions made by neural network models. The article uses a real-world example, the prediction of the net asset values of mutual funds, to investigate the effect of data quality on neural network models. The results of two experiments are reported. The first experiment shows that the error rate (ranging from 25 to 100%) and magnitude of error (5 and 10%) in data used in model prediction affect the predictive accuracy of neural networks. The second experiment shows that the error rate (ranging from 5 to 20%) and the magnitude of error (5 and 10%) in data used to build the model affect the predictive accuracy of neural networks. The findings have managerial implications for users and builders of neural networks working with databases containing errors.