PREDICTING FOREIGN EXCHANGE RATES
About the project
The project is designed to predict foreign exchange rates. It takes opening , maximum and minimum price as inputs and learns and create the model from the trainig data and then you can use that model to predict the closing price of foreign exchange. Training data should be large enough (at least 60,000 sets) in order to predict the output precisely. In our project we are testing our model on a validation set and printing the accuracy as final output.
What we learnt
After completing this project we became well versed with Lua and Torch. Also, one learns the basic concepts involved in the construction of neural networks and the various machine learning algorithms.
How to use
Install torch from this link. Download free forex data from here and convert the csv file into txt file separated by ;. This data will form your training and validation set. Download the code from here and run on your system by using th filename You will also see a graph plot between the expected value and predicted value.
How does the code work
Well, the program uses gradient descent algorithm and uses a test data set to train itself. This is necessary for the program to learn and this model is saved. Now, this saved model is reloaded in a new program and there it predicts the values for the validation set. Then the actual values and obtained values are obtained on the graph.