Predictive Analytics for Product Prices

Abstract

Product prices are affected by company stock prices. If the stock price rises, so will the product price. Otherwise,when the stock value falls, so does the product price. If a company’s stock falls owing to financial difficulties, it may reduce product prices to increase revenues or liquidity. The aim is to estimate whether the product price will rise or fall according to the stock market price. Stock price prediction has risen in prominence due to its function in estimating the future value of company shares. There are various ways to predicting stock prices, including machine learning, deep learning, and ensemble learning. To estimate stock values, the information was collected for a variety of popular companies, such as Amazon. The datasets used are divided into two parts: the first comprises a set of tweets for the stocks under consideration in this study, acquired from the X social media site, and the second includes numerical stock price values. Sentimental characteristics from tweets were retrieved in two various manners to produce polarity. Vader was used to calculate the sentimental score and to generate the percentage of positive, negative, and neutral. Numerical data is also used to generate the sentimental score. All the columns of the two files were merged to obtain one dataset. Then, the problem was framed as a regression task. The evaluation difference between the proposed models was investigated for forecasting stock values according to tweets. In this regard, many ensemble learning models were developed to forecast the price changes of each stock. Furthermore, many machine learning and deep learning models were employed for evaluation. Several assessment indicators were used to assess the effectiveness of the presented techniques. The findings showed that the stacking regressor technique outscored the other approaches, achieving the lowest MAPE in the majority of the datasets.