PREDICTING MARKET SEGMENTS FROM TWITTER DATA USING ARIMA TIME SERIES ANALYSIS

Abstract

Twitter data on social media can be used to predict potential market segments in the future. The continuous nature of Twitter data and data collection sequentially over a certain period uses a text mining process with time series data that matches the actual data. The problem to be solved is to forecast market segments using Twitter data with greater accuracy. Various market segments that are of business interest through the tweets of individual Twitter users have not been utilized optimally. Based on the Twitter data pattern, the research shows that the data pattern is not stationary, so to analyze the data, it is necessary to use the Autoregressive Integrated Moving Average (ARIMA) method. This study aims to analyze time series data from Twitter data and predict market segment predictions using the ARIMA method. The ARIMA method is a method that has advantages in flexibility, the ability to handle stationary and non-stationary data, as well as short-term forecasting with a statistical approach in various time series data forecasting applications. Prediction results using the ARIMA method with an accuracy rate of 94.88%. There are several ways to measure model validation including MAD, MSE, RMSE, F1-Score, and MAPE. In this study, MAD, MSE, and MAPE were used with an accuracy rate of 5.22%. This study succeeded in applying the ARIMA method to time series data from Twitter to forecast market segments with high accuracy, opening opportunities for utilizing Twitter data in business strategy.