Published Paper Details:

The stock exchange is where the shares of listed companies can be bought and sold. The great thing about the stock market is that it predicts that the share price will become a successful trader. Today, the online trading platform offers everyone the opportunity to invest and make money if they accurately forecast the market with their financial knowledge over the past year, for a year. A basic D-mat account has been opened, but this share of traders is profitable. Global indices, demand and supply, news about a company cause stock price fluctuations, and this price fluctuation gives traders the opportunity to make a profit, but the risk is to determine the best price to buy and sell shares. Forecasting techniques play an important role here. Investors predict the market with the help of basic analysis, technical analysis and machine learning technology. Investing in the stock market entails great risks due to nature's uncertainty and instability, which makes it difficult to predict the share price. The internal component of the nonlinear nature and complexity of the measurements makes it difficult to predict. The advantage of an artificial neural network is that non-linear and noisy data can be easily adapted, which improves the input-output ratio for non-linear data. Therefore, share prices can be predicted. Multilevel perceptron (MLP) and the expected retrospective algorithm developed by Rumelhart. This model consists of multi-level programming with three levels of input, output and hidden. Feed Forward means that amount of data go out in one input direction. The MLP process is converted to three levels, the first data sets are loaded into the input neural layer and the processing of the input neurons is sent to the hidden layer and finally to the output neural layer. Each layer of neurons is related to weights in a specific way, the process of changing the weight is called the learning algorithm. This technique involves two types of forward and reverse processes. Forward the uploaded input to the network notes and converts them to output, but the weights must be specified. When we go back, errors are due to the difference between the actual and desired performance of the network, but the weights must be adjusted accordingly.