Predicting Stock Market by Fibonacci Series Approach with Using Various Neural Network Methods

The stock market (SM) is an extremely significant part of the financial industry. Its demand is steadily increasing. SM prediction is the technique of estimating the foreseeable value of a business's stock or other assets that are traded on an exchange for financial instruments. Analysis of the SM is critical for investors and financial companies for making accurate decisions. As historical SM data and Machine Learning (ML) algorithms progress, so does the fascination with employing ML in stock analysis. This study delves deeply into SM analysis using ML with a focus on the use of various ML techniques and approaches. The first step in research is data collecting, which involves gathering historical SM data from sources that includes financial databases, online surveys and APIs. This research focus on stock market of each day prediction based on historical price changes by predictive relationship that exists among Fibonacci series for each organization financial market. These datasets are sourced from Kaggle, and each has 5021 instances from January 2000 to February 2020. Data was scaled, and values were retraced using three Fibonacci percentages, as well as the starting and finishing levels, utilizing several neural network (NN) approaches such as Artificial NN (ANN), Deep NN (DNN), and K-NN. The model's accuracy was measured using the Root-Mean-Square Error (RSME) metric. The RMSE comparison to the dependent variable with lowest value has establish the accuracy level which anticipated being lower in the DNN technique that will be used as the test scenario.