This research focuses on the detection of potential business failures using artificial intelligence methods such as neural techniques and SVMs. The failure prediction model will be an efficient tool for all the actors of the company to detect any potential difficulty. This decision support tool will be built on the basis of a set of financial variables from the literature that reflect the complexity of the default phenomenon. Finally, this research aims to contribute to all decision-makers, internal or external to the company. Indeed, it is a question of proposing a clear vision of the forecasting techniques and of showing the quality of forecasting of the non-parametric techniques as regards the forecasting of the failure of companies. In this context, our research aims at explore a new practical approach based on neural networks and neural networks and SVMs to prevent the failure of firms, in a first step, and failure, in a first step, and to judge on the degree of accuracy the degree of accuracy of the models produced by these two techniques, in a second step. To reach this objective, the prediction quality of the model resulting from the neural network technique is compared to that of the SVM technique.