Prediction of Telecommunication Company Stock Price using Multiple Linear Regression

Abstract

This study presents case studies in finance, mainly stock trading. The paperwork from a portfolio data mining effort and data obtained directly from the Indonesian Stock Exchange serves as the basis for the case study. A technique for assessing future occurrences based on relational patterns in data is prediction (forecasting). This technique is used in field data mining on time-series data that is numerical in characters, such as dates or attribute dates. Examples of prediction techniques include anticipating the temperature, weather, foreign exchange rates, and stock prices for the upcoming three months, determining whether the speed limit will be increased, and more. The "close" variable is the target variable. At the same time, the "prior, open price, first trade, high, low, index individual, offer, bid" variables are predictive variables that can be used to extract information that can improve the accuracy of the forecast value—the stock prediction outcomes of PT. Telkom Indonesia Tbk (TLKM) is predicted using the data mining model developed using time series data. Investors can use this information to decide what to do or establish procedures to prevent losses while buying and selling stocks. Testing R2, AUC, MAE, MSE, RMSE, and MAPE determine the accuracy of the stock prediction results. The experimental results from the best MLR model after experimenting with feature engineering yielded an R2 value of 100%, an RMSE value of 0.0548, and a MAPE value of 0.0013%.