#TITLE_ALTERNATIVE#

• Main Author: (NIM : 10213032), JAMALUDIN
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/22669
• Institution: Institut Teknologi Bandung

Stock market is an example of complex system which has became object study in physics. From the previous study we knew that to understand about the system that underlying in stock market, it could be looked by the correlation between the stocks within the system. The correlation that exists between stocks might not be stationary because market condition always changes by the time. As a result random contribution (noise properties) would be found in cross correlation of stock data. Random Matrix Theory (RMT) was the method used in this final project to filter noise properties from stock data by comparing the characteristic of eigenvalue and eigenvector of stock correlation matrix by applying random correlation matrix. After stock data has been cleaned from noise, we were going to analyze group within stock market using complex networks (CN) and block diagonal matrix (BDM) approach. In complex networks approach, stocks can be treated as a nodes and correlation between them as its edges. Advanced Label Propagation Algorithm (LPAm+) was algortihm used to identify groups in stock market network. Basically LPAm+ will determine group of each node based on the most frequent label of its <br /> <br /> <br /> neighbors. Then in block diagonal matrix approach, the correlation matrix of stock data was converted to block diagonal matrix which each block represented a group in stock market. Simulated annealing (SA) was algorithm used to construct block diagonal matrix. Simulated annealing algorithm is used to find global minimum of an objective function which is similar with concept of annealing for constructing crystal. For this final project, 502 stocks data have been collected from January 1, 2007 until October 28, 2016 from S&P 500.