High Frequency Trading Prediction

Intro

In this project, we explore several machine learning techniques in order to predict the raise and fall of trading price.

Data

High frequency trading records from April 2016 to May 2016. Only fifteen futures are included. Omitted in this repo, sorry for that :(

Methods

Efforts involve several ways to transform raw records into trainable inputs. Models used include traditional ones, like SVM, LR and also neural networks like baisc MLP and RNNs.

Also, we do simulated tradings to evaluate the performance of models, and yields satisfying results. ( Maybe we failed to consider some situations, the results are too good to be true...

Code

Core models are implemented by Keras, thanks for the convenience :)