The study deals with a very important aspect of a country, the Stock Market (SM), which is the financial backbone of an economy. Nowadays, on average, 1/4th population is associated with the stock market, directly or indirectly, through mutual funds. The research integrates traditional technical analysis with cutting-edge reinforcement learning to improve the trading experiences of the traders and generate more returns. The proposed study is based on Simple Moving Average (SMA) crossover along with Average Traded Volume (ATV) and Relative Strength Index (RSI) confirmations of SMA crossover signals. SMA is the main signal generator that generates buy and sell signals during Golden Crossover (GC) and Death Crossover (DC), respectively. When this GC or DC appears with rising volume, ATV in the upwards direction confirms buy or sell signals. There is a very low probability of sustaining the GC and DC signals without high volume. Further, the signals are verified with RSI development to improve the market prediction level. These multi-indicator-based signals are executed through a Reinforcement Learning (RL) based algorithm to train the system in different market conditions to generate maximum profit. Back testing is done on 5 years of historical data of 10 large-cap stocks listed in the National Stock Exchange (NSE) in different time frames to compare the accuracy, Return on Investment (RoI), Sharpe ratio and Sortino ratio. The model was able to generate 85.00%, 78.35%,71.27 % accurate signals in daily, hourly and five-minute time frames with 260%,29%,4.92% RoI in 5 years, 1 year and 1 month, respectively. The study focuses on the integration of smart rule-based indicators with an advanced reinforcement learning algorithm for automation.
Kadia et al. (Thu,) studied this question.