Need Help With Stock Predictions

Part 1 (2017)
1

I’m a new Deep Learner trying to do Stock Predictions, but in my own way - not watching any YouTube videos that show working methods to doing this. I need help doing it my way, though :smiley:

I created a method to get the information of any Stock, day-by-day, for up to 20 years. For example, I could get Tesla Inc. (TSLA)'s stock info in a csv like this:

Screen Shot 2017-07-06 at 4.10.31 PM.png1180×378 50.3 KB

Now, here’s where it gets complicated (for me). You take the datapoints (open, high, low, etc.) for the past 20 years (X) to get the next day’s data (Y). Then you repeat this for the thousands of Stocks’ CSV data I have.

IMG_0118.JPG2448×3264 2.93 MB

In my mind, that’s how you would have to go around to train this. But it seems kind of weird. It’s like training the model a bunch of times, each time for a different stock. Is that what I’m supposed to do?

Hopefully you understood me! Thanks :slight_smile:

2

Jeremy discussed Time Series analysis in lesson 14. You should go through that lecture.

3

As you say, you are the beginner in DL. In this case, lecture N14 is not your choice, at least not yet. I can advise you to start the course with the first lecture and try a simpler approach now. You can split the data for each stock for a fixed period of time (for example, 30 days) and try to apply the CNN network. In this case, I believe that your network can learn about the [MACD (Moving Average) Convergence / Divergence Oscillator)] (http://stockcharts.com/school/doku.php?id=chart_school:technical_indicators:moving_average_convergence_divergence_macd) and many other technical indicators.