About the Neurolyser Project The Neurolyser project initially started under the name `Smart Stock Manager'. Therefore in the first version on this software the initial splash screen flashes the name Smart Stock Manager. This was developed as a part of a semester mini-project. The software which is the outcome of this project has been developed using Java basically due to the platform independence nature of Java.

The system uses Artificial Neural Networks - Error Back Propagation Algorithm. It consists of a ‘Multilayered Feed-Forward’ circuit. There are mainly 3 layers. First is the Input Layer in which the number of nodes varies according to the accuracy desired by the user. The next layer or the Hidden Layer may consist of a fixed number of nodes. The system is designed to predict for the next 7 days. Hence the number of Output Layer nodes is also 7. The Activation Function chosen is a Sigmoid function. Every edge in the signal path is associated with a Synaptic Weight, represented here by two matrices. Since the model here is a non-linear one, the number of nodes in each layer varies. The input array consists of prices of a particular stock along with a set of four constraints on a scale of 10.

These values are then used to make the prediction. (For more details see the Documentation Page)