Correctly predicting price movements in stock markets carries obvious economical benefits. The task is traditionally solved by analyzing the underlying company, or the historical price development of the company s stock. A third option that is undergoing active research is to create a predictive model of the stock using machine learning. This thesis follows the latter approach, in which a machine learning algorithm is presented with historical stock data. The algorithm uses this information to train a model that is expected to infer future prices given recent price information. Machine learning is a large field within computer science, and is under constant development. Breakthroughs in a family of machine learning models known as artificial neural networks have spiked an increased interest in these models, including applying them for financial prediction. With a plethora of models available, selecting between them is difficult, especially considering the constant flow of emerging models and learning techniques. This study compares a selection of artificial neural networks when applied for stock market price prediction. The networks are selected to be relevant to the problem, and aim at covering recent advances in the field of artificial neural networks. The networks considered include: Feed forward neural networks, echo state networks, conditional restricted Boltzmann machines, time-delay neural networks and convolutional neural networks. These models are also compared to another type of machine learning algorithm, known as support vector machines. The models are trained on daily stock exchange data, to make short-term predictions for one day and two days ahead. Performance is evaluated in the context of following the models directly in a financial strategy, trading every prediction they make. Additional performance measures are also considered, to make the comparison as informed as possible. Possibly due to the noisy nature of stock data, the results are slightly inconsistent between different data sets. If performance is averaged across data sets, the feed forward network generates most profit during the three year test period: 23.13% and 30.43% for single-step and double- step prediction, respectively. Convolutional networks get close to the feed forward in terms of profitability, but are found unreliable due to their unreasonable bias towards predicting positive price changes. The support vector machine delivered average profits of 17.28% for single-step and 11.30% for double-step. Low profits or large deviations were observed for the other models.