TEXT MINING FOR DATA TIME SERIES IHSG STOCK

• Main Author: Khairunnisaa, Kurnia
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/42374
• Institution: Institut Teknologi Bandung

The purpose of this research is to determine the stock price prediction model approach based on internet news. Text mining is used in collecting, processing, analyzing, and gaining insights from internet data. Internet news information insights were collected in the form of words that represent the news. Stock prices information insights were collected using 1 day IHSG closing stock price gap and 7 day IHSG closing stock price gap. The collecting of internet news were narrowed to sources, topics, and additional informations from the global news. The collection of internet data were performed on “X1” and “X2” as sources with “Y1” and “Y2” as topics. Data matrix were built on the information insights of the internet news and the IHSG closing stock price gap. Multicollinearity assumption test and KMO from the data matrix are to be fulfilled before performing the cluster analysis. K-means and average linkage clustering were performed on the data matrix in order to find the cluster similarities based on the news distribution each day within fixed time period. Cluster analysis were performed to observe the stock price fluctuation from the news distribution provided by the cluster. K-means clustering yields the better accuracy compared to the result of average linkage clustering.