For many years, natural gas prices were strongly correlated with those of crude oil. Recently, natural gas prices started to show an independent trend. Natural gas prices are driven by the law of supply and demand which is reflected by the weather and inventory levels among other factors. In the last decade, electronic trading platforms took over the exchanges. With the advent of algorithm trading (AT) and in particular high-frequency trading (HFT), trading commodities, which include energy trading, became riskier due to their extremely volatile nature. This dissertation presents a novel framework that provides insight into the use of HFT in natural gas futures markets. Since there are no publicly disclosed data on such practices, the objective is to develop a comprehensive model for natural gas futures trading. A new heuristic simulation, predictive modeling and optimization algorithm that automates trading natural gas futures is proposed and evaluated. Simulation is used to reconstruct the order book using top of the book natural gas futures historical data. Predictive modeling techniques based on multi-class support vector machines are used to predict the occurrence and the amplitude of spread crossings. Finally, an inventory optimization model is used to determine optimal trading volumes for each trading period. Two types of trading strategies are derived: a strategy using Immediate-Or-Cancel orders where an order is totally or partially executed while the remaining is cancelled, and a strategy that limits orders’ cancellation. Both strategies are tested with real and synthetic data. In this setting, both strategies can lead to profit. This could be used by policymakers and market regulators to implement order cancellation restrictions on commodity futures trading to prevent harmful speculation.