Abstract:
"
Stock market is a crucial factor which contributes to the development of a country. The reason
for considering this statement is because of the transactions that happen in the trading time
brings out huge capital gain from investors and traders via trading companies. Therefore the
transactions of investors and stock traders are very important to keep the market alive.
Considering the stock traders in the short-term market analyzes the performance of a company
by the past values of those company index. However analyzing those values especially closing
price for the trading is not sufficient for them get a good trade. There are many existing systems
which forecast the prices for a long time in future but due to high volatile nature those values
cannot be always accurate. Therefore limitation of analyzing both past transactions and
impacting features to predict the close price of a day brings out the main research problem to
be addressed in this project.
This proposed solution uses ensembles method of machine learning models such as Random
Forest, XGBoost, SVM, Decision trees and Lasso to predict the next two days of the closing
price. Also an additional supportive feature analyses news sentiment which affects the pattern
of stock price and prompts suggestion for traders. The evaluated system’s accuracy was
measure with RMSE, MDA, MSE and the overall accuracy of news sentiment model gave
54% where the whole system was efficient and satisfactory when benchmarking with existing
systems. Furthermore the system was evaluated by domain, industry experts and end users
with well-designed evaluation criteria"
