How to take prediction output?

Quant Trading Strategies & Ideas
1

Course Name: Deep Reinforcement Learning in Trading, Section No: 13, Unit No: 6, Unit type: Video



I understand that it buying/selling in the backtest function, but how to see where it has predicted buy/sell. I would like to manually take output trading.            

 

2

Hi Mahesh,



The "pos" in the backtest is the position being taken. -1 for sell, 1 for buy, 0 for square-off. 



For live trading, the template code shows the action (E.g. "Sell action encountered") taken in the window as can be seen below:







If you don't want to trade, you can comment the line which is taking positions.

 

3

Hi Gaurav,



Thanks for the reply.



I would like to know taking these predictions is possible row wise (date/time wise) in a data frame to be able to see whether it predicted buy/sell for a particular row.<br />


For example, with a particular conditions of market in a row, what is its prediction buy/sell.

4

You can construct a new state at each row, and use the q_network to predict the action for the given state. This gives you the action row-wise.



Hope this helps.

5

I am not a seasoned coder,

it would be really helpful if you can please show me a sample code on how to do that.

6

You can refer to the live trading template for the sample code. The code inside the strategy function will need to be put in a for loop if you are running on a historical data frame. The bars5m data would be the rolling window of historical data needed to create the state.

 

7

Thanks Gaurav, will try it out.

8

You can refer to the following notebook: Quantra by QuantInsti | Courses on Algorithmic and Quantitative Trading



In this notebook, the dataframe trade_analytics_df will help you with the positions (Buy, Sell) take by the RL in the "position" column and at what timestamp (Time index). A sample output from trade_analytics_df is shown below.



The value of +1 indicates Buy signal, -1 indicates Sell signal and 0 indicates no position.



You can use this position column with your other trading signals and trade using combined signal.



I hope this helps.





9

Thanks Ishan.



This helps.