BUILDING TRAJECTORY DATA WAREHOUSE TO SUPPORT TRAJECTORY CLUSTERING

• Main Author: Logianto, Albert
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/36506
• Institution: Institut Teknologi Bandung

The development of tools that can track the location of objects such as GPS causes the number of spatio-temporal data to continue to increase. Spatio-temporal data is data that has attributes of space and time. Moving objects data is one example of spatio-temporal data. Examples of moving objects data are taxi movements in a city and animal migration movements. The growth of the number of spatio-temporal data is also followed by the need to analyze the data. One analysis that can be done is trajectory data mining. There are several types of tasks in trajectory data mining, one of which is trajectory clustering. Trajectory data warehouses can be used to support task clustering tasks. However, there are currently not many data warehouses that can support this specific requirement. In this final project, a trajectory data warehouse system is developed to support trajectory clustering, especially those that utilize the TRACLUS algorithm. Trajectory data warehouse helps the clustering process in handling several problems such as filtering noise and constructing trajectories data from raw data. By using data from the trajectory data warehouse, the trajectory clustering system no longer needs to do pre-processing to eliminate noise and process raw data to form trajectory data. The tools used for the ETL process are Pentaho Kettle and PostGIS is used for the data storage system. Tests are carried out using the TRACLUS trajectory clustering algorithm for taxi GPS data in Beijing. Testing shows that the data warehouse trajectory system can produce the valid data and is suitable to be used by trajectory clustering tools. This Trajectory data warehouse can also be accessed to generate various queries that are useful for exploration of the trajectory data.