PREDICTION OF STOCK VALUE MOVEMENT USING TECHNICAL ANALYSIS AND SENTIMENT ANALYSIS OF HEADLINES NEWS

• Main Author: KURNIAWAN MUTTAQIN NIM: 23514072, HANNIF
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/22355
• Institution: Institut Teknologi Bandung

Stock price movements have random fluctuations that are influenced by <br /> <br /> speculative factors that are difficult to predict accurately. The behavior and <br /> <br /> opinions of investors have an influence on stock price movements. The time series <br /> <br /> model can be used to solve the problem. <br /> <br /> This study uses technical analysis features from historical stock prices, and <br /> <br /> sentiment analysis features from headlines news that can be used to create <br /> <br /> forecasting models. Multiple Kernel Learning (MKL) is used because the data <br /> <br /> source comes from different sources and has different data distribution. The MKL <br /> <br /> method can improve the effectiveness of choosing the kernel, it can minimizing the <br /> <br /> error value because uses less precise kernel. <br /> <br /> The feature extraction results against the data source generate eleven features. <br /> <br /> These eleven features can be used in making predictive models using the MKL <br /> <br /> method. The sentiment analysis feature can improve the best MAE accuracy by <br /> <br /> 24.72% in the MKL method, and 39.84% in the SVR method.