This article proposes a Neuro-Genetic model for 1-day and 7-day stock price predictions. This model is the Back-propagation Artificial Neural Network with the number of nodes in a hidden layer optimized by Genetic Algorithm. The number of input nodes is the result of autocorrelation analysis of communication stocks of the Stock Exchange of Thailand. The success of this model is based on two performance measures, hit rate and realised potential. This research used this model to predict stock prices compared with the Naïve prediction model. The result shows that both average hit rate and realised potential of the model are larger than of the naïve prediction model.