AutoML Trading: A Rule-Based Model to Predict the Bull and Bearish Market

Many researchers from different fields have tried to figure out how to forecast the future and trade on the share market since it is hard to accomplish because the market is so complicated. Machine-learning techniques have been exceptionally productive in this domain in recent times. Financial experts are creating a “trading system” for stock returns and trend forecasting. Since “trading online” was introduced, the amount of transactions that take place on the equity market each day has seen a significant increase. The objective of this study is to propose a rule-based trading framework not only forecasts the market but also finds out the best bull and bearish weekdays for trading. So here in this framework, we proposed an automated machine-learning (AutoML) online trading system that combines two modules. The first module finds the best weekdays to be bull or bearish with an algorithmic trading strategy. The second module combines two sub-models: an ensemble-based classification module and an ensemble-based regression module. The bagging classifier with the Extra Tree classifier is used as the classification model, whereas the bagging regressor with SGD regressor is used for the regression model. Different matrices were followed as per the criteria with Nifty-50 indices for model performance evaluation purposes by us. The regression model has an average RMSE value of 0.2091 MSE value of 0.0551, and the classification model’s accuracy range varies between 95.65 and 99.59. With the help of this framework, even if a little knowledge, investors can minimize investment loss and maximize their profit.