rl_generate_policy function

Function performs Reinforcement Learning using the past data to generate model policy

This function will perform Reinforcement Learning using Trading Data. It will suggest whether or not it is better to keep using trading systems or not. Function is just using results of the past performance to generate the recommendation (not a holy grail).

rl_generate_policy(x, states, actions, control)

Arguments

• x: * Dataframe containing trading data
• states: * Character vector, Selected states of the System
• actions: * Character vector, Selected actions executed under environment
• control: * List, control parameters as defined in the Reinforcement Learning Package

Returns

Function returns data frame with reinforcement learning model policy

Details

Initial policy is generated using a dummy zero values. This way function starts working directly from the first observation. However policy 'ON' value will only be generated once the Q value is greater than zero

Examples

library(dplyr)
  library(ReinforcementLearning)
  library(magrittr)
  library(lazytrade)

  data(data_trades)
  states <- c("tradewin", "tradeloss")
  actions <- c("ON", "OFF")
  control <- list(alpha = 0.7, gamma = 0.3, epsilon = 0.1)
  rl_generate_policy(x = data_trades,
                     states, actions, control)

Author(s)

(C) 2019,2021 Vladimir Zhbanko