In this research different innovative deep learning approaches for predicting the future values of time series are compared – the standard classification of a time series, used for predicting the trend of the time series and the forecasting of future values. The reasoning behind the research is to find the best tools for trend prediction. Both state of the art as well as new datasets are used to generate the results. These results are compared using both standard and new measures, designed to assess practical applications of the analyzed deep learning approaches in trend prediction. The presented results give strong foundation for future work, especially in terms of merging various approaches into one, boosted solution.