TREE-BASED MODELS FOR PREDICTING DIGITAL BANK STOCK VALUATION USING MULTISOURCE INTERNET DATA APPROACH

• Main Author: Mangaratua, Lucas
• Format: Theses
• Language: Indonesia
• Subjects: Manajemen umum
• Online Access: https://digilib.itb.ac.id/gdl/view/65063
• Institution: Institut Teknologi Bandung

Digitalization has been an important subject in the last decade, especially in Indonesia since the COVID-19 pandemic effect. The Indonesian government is expected to support the building and improving the development of the digital economy. Also, the market is estimated to respond well to digital payment for the next 2 years. The three giants of the digital bank in Indonesia are Jenius, Jago, and Neobank have their own competitive advantages in the market: Jenius with the best brand awareness and highest annual net income, Jago with being the top 6 market capitalization on the Indonesia Stock Exchange (IDX), and Neobank with the most applications’ total downloads. By knowing how important the digital bank is in the future stock market, a study to predict the digital bank company stock price is necessary to develop. Utilizing various internet data such as Google Play Store reviews, Twitter, and Google Trends, a study case of digital bank company stock price prediction is developed. The model prediction is build using a machine learning tools such as Regression Tree and Random Forests model as part of tree-based models. Hence, based on the Regression Tree and Random Forests variable importance calculation, the Google Play Store review and Google Trends data are the most relevant to the model prediction. The information from these 2 sources generates more contribution to the model than Twitter does. Based on the stock price prediction, using Regression Tree and Random Forests, the forecast cannot make an accurate prediction for the portfolio performance of Bank BTPN using the Jenius model prediction, but success in accurately predicting the portfolio performance of Bank Jago and Bank Neo Commerce using the Jago and neobank model prediction.