E-commerce is growing rapidly and is constantly gaining momentum towards being the dominant source of commercial transactions. The pricing policies and pricing strategies of businesses are of paramount importance for surviving in this highly competitive market, achieving sell-out goals and maximizing profits. Towards this end, various dynamic pricing algorithms have been proposed and adapted to the continuously changing conditions of online markets. These algorithms are based on the abundance of data available to the online stores about market conditions as well as customer’s preferences and consumption habits. Effectively analyzing this data and being able to integrate them into dynamic pricing strategies can give a significant competitive advantage to businesses. The purpose of this thesis is the development of a system for dynamic pricing of products of e commerce stores. We proposed an improved hybrid model that is used to solve the univariate timeseries predictions problem, in order to predict future sales. The proposed model uses a deep neural network (LSTM), which has shown promising results in the lasts years compared to classic feedforward neural networks. Moreover, we proposed an optimization algorithm for product pricing that optimizes the conversion rate and the profit margins of e-commerce stores. Finally, we evaluated our system be creating a simulated marketplace using real, anonymous data.