INTEGRATING CLUSTERING ANALYSIS INTO ARIMA ANDSARIMA MODELS FOR PREDICTING STOCK PRICEMOVEMENT PATTERNS OF IDX30 INDEX

• Main Author: Ilham, Muhammad
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/82744
• Institution: Institut Teknologi Bandung

Stock investments offer various attractive benefits, such as high potential returns, good liquidity, and the opportunity to participate in corporate growth. However, stock investments also carry risks, including stock price fluctuations and market risk. Portfolio diversification is a strategy used to reduce risk by investing in a variety of assets. This research aims to accelerate the diversification process by grouping IDX30 stocks into several clusters based on historical patterns using the K-means algorithm. Considering that the goal of investment is to gain future profits, the average stock price of each cluster is predicted for several days ahead using ARIMA and SARIMA time series models. The results show that IDX30 stocks can be clustered into three groups with upward, downward, and stable trends. The portfolio from the group with an upward trend has the highest Sharpe ratio of 0.0449, while the group with a downward trend has the lowest Sharpe ratio of -0.811. The Sharpe ratio measures portfolio performance relative to its risk; the higher the Sharpe ratio, the better the portfolio performance. ARIMA and SARIMA models produced fitting results with MAPE below 50% and prediction results with MAPE below 25%. MAPE measures the accuracy of the predictive model by comparing predicted values with actual values; a value below 50% indicates an acceptable level of error.