FORECASTING THE NUMBER OF VISITORS USING HOLT’S EXPONENTIAL SMOOTHING WITH CCTV ANALYTIC FOR DATA COLLECTION (CASE STUDY : BANDUNG STATION)

• Main Author: Hanavi
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/64375
• Institution: Institut Teknologi Bandung

The high societies interaction in public area or other locations is a serious challenge for improving the surveillance system. The Close Circuit Television (CCTV) surveillance system is expected to be a solution, but in fact it has not contributed. CCTV which is passive monitoring and does not provide feedback needs improvement to become intelligent monitoring capable of recognizing objects and counting the number of people. Even though so many research about people counting, but the issue of counting the same object repeatedly in waiting room or service location (slow moving object) is still a problem and how to use the data that has been collected remains unanswered. Using the stages of Design Science Research Methodology (DSRM), research produces artifacts to solve problems through the stages of designing artifacts, demonstrating artifacts and evaluating artifacts. Artifacts in the form of methods and models answer the research objectives to minimize the number of the same object repeatedly and analyze the dataset counting to see the trend of the number of visitors based on case study data from Bandung station. Holt’s exponential smoothing forecasting method was choosen because the characteristic of counting dataset are time series, fluktuatif and adjusted to the research objectives. The artifact results of the lost centroid handling method succeeded in maximizing the accuracy of counting from 42% to 72% in slow moving locations. The average accuracy at the five Bandung station locations is 85.4% with the highest 94% at the close-range south exit location and the lowest at the north entrance waiting room location at 73%. The results of artifact modeling forecasting Holt's exponential smoothing succeeded in predicting trend the number of visitors with the lowest RMSE value of 22.12 at the north entrance gate.