Presenter: Dr. Yves Hilpisch of The Python Quants     

Length: 18 mins
It is easy to retrieve historical intraday data via the Eikon Data API, with Plotly and Cufflinks making visualization of the data simple. Deep learning techniques such as deep neural networks (DNN) for classification are easily applied via Python, and these techniques may be helpful in predicting the direction of market movements. Deep neural networks are increasingly prevalent in financial markets, and we explore how they can be applied to predict financial time series.

• Retrieving historical and intraday data across asset classes with the Eikon Data API, and working with this using Pandas, Plotly and Cufflinks
• Preparing lagged data on which to train the network
• Implementing the Deep Neural Network (DNN) model and training it to the data. A classification algorithm that is able to learn from historical patterns to predict whether an up or downwards movement is more likely. We use the Tensorflow package here

• Testing the predictions generated from the model using vectorized backtesting. We then conduct a train-test split and apply out-of-sample testing, to build a more realistic picture.