A Tech Talk Picking Stocks with Google Cloud Platform and Tensorflow Machine Learning

I had a wonderful evening speaking at the Google Cloud Platform Usergroup Singapore November 17, 2017. Thank you very much to Google and Huiyi for the invite and facilitating a great event.

Thanks for letting me share my work on Stock Market Prediction with Datalab.

I shared aspects of my ‘Follow the Sun’ work with the Google Cloud Platform, using Keras and Tensorflow. I will be sharing more details and my notebook in the future, but I was glad to have a nice discussion with so many like-minded people.

As a champion of actionable intelligence, it has always been my stance that neural networks are not a magic bullet, and that technology and techniques do not solve the problem by themselves; better data is the solution and domain knowledge is essential!

Using the Playground simulation, I was able to show that increasing the  computing power and hidden layers in the network improved performance, but not nearly as much improvement as by simply adding additional features to the data. Try it out for yourselves; you can play with a simulation here:

The audience sat through the painstaking preparation of the data, incorporating time series data about related stocks and competitors, and the final output -- but I’ll only share the results in person, so comment or reach out if you’d like to find out.

I see my role as helping people, and as a teacher, sharing new ideas and answering questions.

You can also refer to this excellent solution from Google's Corrie Elston about machine learning with time series.