In the current scenario, one of the recent research works is stock prediction in the future using machine learning techniques. Today, the stock market is one of the greatest investments to retail their current shares to get profit. This necessity forces everyone to turn back to predict the future stock price market using the efficient techniques in machine learning. However, the latest market analysis and prediction is one of the most complicated tasks to decide the value. This article proposed stock price prediction using an improved algorithm in machine learning. Machine learning uses different types of models to identify the prediction easier and accurately. This chapter presents the use of LSTM based machine learning to predict stocks and factors considered are date, time, closing price and opening price of the stocks. Finally, this machine learning model was trained with numerous data, and the results were compared with existing algorithms. These performance results show the maximum prediction accuracies of 92.10% and 84.10% were attained using improved LSTM model.
TopIntroduction
The stock market plays a vital role in Indian economy. Today, the Indian economy highly depends on Stock market and agriculture. Agriculture and stock market is one of the back bones of today’s income to Indian economy (Andersen, 2018) . It is one of the most significant investment opportunities for businesses and investors. When a firm expands its business through an Initial Public Offering, it can make a lot of money. It's a good moment to start something new an investor who wants to buy fresh stocks and make a profit dividend paid out as part of the company's bonus scheme for shareholders. An investor can also trade stocks as a trader in the stock exchange. Stock traders must be able to foresee stock market movements in order to make the best decisions on whether to sell, hold, or acquire other stocks. Stock traders who want to make money should acquire equities between raise and fall of stock market. Stock traders that properly predict stock price patterns can make significant profits. As a result, stock traders need to be able to foresee future stock market movements in order to make informed decisions. Investment in the stock market is dangerous, but when done correctly, it can be one of the most cost-effective ways to earn big profits (Altay E. &., 2005). To prevent buying dangerous stocks, investors assess a company's performance before deciding to purchase its stock. This assessment includes a look at how the company performs on social media and on financial news websites. Investors, on the other hand, cannot fully examine such many social media and financial news data. As a result, investors will need an automated decision support system, as this system will evaluate stock movements automatically utilising such enormous amounts of data. Machine learning techniques can be used to create this automated system (Nigussie, 2017) (Porshnev A. R., 2013). Artificial Intelligence research have taken a particular interest in this field since precise stock prediction based on external factors will boost investors' profits (Creamer, 2007) (Jammalamadaka, 2019).
The complexities of stock market prediction are anticipated with machine learning algorithms. Machine Learning, as stated in the introduction, requires the user to give a target values and label must be anticipated, as well as independent variables can be to identify the target variable’s values (Xiao, 2014) (Obthong, 2020) . Machine learning is distinct from conventional predictive models. It employs optimization algorithms, cross-validation procedures, complex mathematical algorithms, and other advanced computing techniques to arrive at the result, which is highly accurate (but low in interpretability). Stock declaration made in advance that happen correct might influence meaningful rewards for traders. Prediction exists commonly recognised to be troublesome alternatively random, that mean possibly predicted by painstakingly examine the history of the appropriate stock exchange (Marcek, 2014). Machine learning is a good method to express these types of movement. It forecasts a package and sell goods value namely about the tangible value, reconstructing accuracy. Many human beings bear benefited from the request of image processing, machine intelligence to the field of stock declaration made in advance because of allure adept and accurate measures (Fischer, 2018).
By automatically pulling data from the higher dimension, a Deep Neural Network (DNN) can represent the lower dimension. Hierarchical neural networks are designed for integrating the bias of the respondent. Reinforcement learning does not express perception fully despite its use in decision-making processes (H. Srivastava, 2017).In order to achieve this, we have integrated Deep Learning with Reinforcement Learning because each of these methods complements the other (Selvin, 2017) (Mehtab, December 2019). A cognitive decision-making system of the sophisticated system can be created using an integrated approach. As a result, Deep Learning has excellent perception and feature extraction abilities (Murkute, 2015). The proposed work has three stages: Initial, Internal and Final Stage.