Following changes can be incorporated which will increase the dependability of this method:
The no of days for which the data was taken can be more, more so when it has been mentioned by Simon Haykin that “the appropriate no of training examples is directly proportional to the no of weights in the network and inversely proportional to the accuracy parameter (i.e. performance goal: MSE)”.
The input parameters taken for this simulation are adequate for this forecasting, but an improved condition is when more related parameters like the Earning per Share (EPS), PE-ratio (Price to earning ratio), etc are taken.
More advanced algorithms in Back-propagation can be employed.
Specific tools like the industry standard neural network/adaptive system simulator NeuroSolutions can be used for greater insights.
 H. White, “Learning in neural networks: A statistical perspective”, Page 425-464, Neural Computat . 4, 1989.
 T. Masters, “Practical Neural Network Recipes in C + +” , Academic Press, New York, 1993.
 Tan Clarence N W, ‘An Artificial Neural Networks Primer with Financial Applications Examples in Financial Distress Predictions and Foreign Exchange Hybrid Trading System’, School of Information Technology, Bond University, Australia
,  Kaastra Iebeling, Boyd Milton, ‘Designing a neural network for forecasting financial and economic time series”, Page 215-236, Neurocomputing 10, 1996,
 MATLAB 7.0 Help Files
 Refenes AN et al, “Stock Ranking:Neural Networks Vs Multiple Linear Regression”,Department of Computer Science, University College of London, UK