When faced with the struggle to extract insights from complex and noisy data, often the end user may assume that there exist no significant relation between the features and target in the dataset and is forced to either quit the study or resort to alternate means. Artificial Neural Networks (ANNs) might be of help to predict some of the most complex data used in the industry. But it is neither easy to identify such situations where ANNs might be more useful as compared to some of the other techniques, nor to identify the best Artificial Neural Network(ANN) to be used in a given situation for the available data. This paper deals with comparing the two most popular variants of Artificial Neural Networks, namely the Multi-Layer Perceptron (MLP) and the Radial Basis Function (RBF) Neural Networks with respect to their capabilities to predict noisy, complex, heteroskadastic or multi/ mixed-distribution data. The data used for the study has been synthetically produced and is subjected to complex transformations, correlations and random noises of different orders so that it can best represent and surpass the complexity of average data available in any industry. Therefore the results of the study may be used in any industry, over most of its complex data scenarios. At the end a comparison of the results of applying these two ANNs and some of the other popular statistical prediction algorithms is produced to identify the best algorithm for predicting the most complex data.