This paper presents research on a profitable trading strategy for G10 currencies. We will devise trading strategies by considering realistic trading scenarios, analyze the performance of such strategies on out of sample data, identify the risks of these trading strategies, explain why the trading strategy works, and summarize and draw conclusions. We will use technical indicators like Moving average (MA), Exponentially-weighted moving average, Ichimoku, Relative strength index, stochastic oscillators, Williams %R, Commodity Channel Index, and Bollinger bands. We will also use fundamental indicators like interest rate differentials from one currency to another. We will build a machine learning based model as it would be better equipped to build dynamic trading rules to capture profitable trading opportunities. We used an ensemble of machine learning algorithms - logistic regression with four principal components, and a voter classifier with random forest, extremely randomized trees, logistic regression with 4 principal components from the features, and support vector machine with 4 principal components from the features, giving 27% and 27% annualized returns, and 1.47 and 1.60 Sharpe ratio respectively.