Stock index data are highly volatile in nature and changes over a period of time. Prediction of these are really challenging and computational models are very useful. A computational model called PLS-Neural network has been used in this study. To explore stock market tendency the closing prices with thirteen variables are considered from te BSE sensex data. To evaluate the prediction ability of the models, standard error values are calculated. PLS regression together with Neural network gave a good prediction using a common error measure.