Power generation from tidal currents is emerging as a major type of renewable energy. It is estimated that up to 5,000 GW of power can be extracted from the world's oceans. One of the challenges to optimize the productivity of underwater turbines is knowing where to place them. Existing data on ocean flow are discontinuous and cannot be easily used to describe or predict current patterns across all waters. In this project, we created a machine learning framework to predict water velocity for prospective turbine sites. An interactive web interface was also implemented to help users make sense of the velocity, power output, and cost information and explore options for the development of marine turbine systems. Using historical ocean data from major US agencies such as NOAA and Scripps Institute of Oceanography, we collected information on current speed, salinity, and depth to create our feature set. After trialing several models, we took one supervised approach to machine learning, decision-tree regression, an algorithm built into the scikit-learn library. The resulting model was able to accurately predict current speed for the ocean test set. The web app, written in Flask, presents an economic analysis based on the current velocity predictions. Data is then visualized using the Google Maps API, which generates a heatmap indicating which locations are suitable for our prospective users in siting marine turbines for power generation. We used 5-fold cross-validation in the training process to select the best hyperparameters, and the optimization resulted in a training R2 accuracy score of 0.88 and a testing R2 score of 0.75, with the training data covering a substantial portion of North American coastal waters. To help investors and city planners make decisions on developing ocean current energy, the web app also provides financial analysis that includes production costs, annual revenue, energy transmission costs, and payback time. In the map page, the user can view and search for locations within 200 km of North American coasts. They can then narrow down their search with selectors such as month, time, depth, number of turbines, etc., and then click on a point to learn more about that specific area. Based on our predictions, there are many potential sites to implement ocean turbine energy at a cost-effective power level near North America. Our framework could also be scaled and applied to datasets in other locations of the globe.