ONLINE CHANGEPOINT DETECTION FOR PREDICTING WEATHER CHANGES

• Main Author: B Y Silitonga, Matheus
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/36422
• Institution: Institut Teknologi Bandung

Detection of changes in a process is a very important topic in wide variety of fields, such as finance, biology, law, and even used for detecting weather changes that are often not realized by ourselves. If we have a time series where the process changes can be seen clearly, then the histogram of this time series can clearly show changes in the distribution of the data. However, the problem is that often the change in a process cannot be seen clearly in the data. So, in the histogram of the data itself, changes in distribution cannot be seen clearly. For this reason, an effective method is needed to sort time series data until we find the division of the historical data into several regimes that represent changes in distribution at each regime change. The method stated in this thesis is an online changepoint detection. Changpoint detection is a recursive method for detecting sudden changes in generative parameters of a time series. This changpoint online detection is a method used to see patterns found in an observation data. With the detection of changes in patterns found in each regime of observation data, we hope to see pattern at the end of historical data from observational data in order to predict the appearance of datums in the future. For this, the author conducted a simulation to find out first the application of the online method (EXO) in simulation data. From several studies, it was seen that three of the four original points were detected using the EXO method by using the hyperparameter values that were searched using numerical methods. At the end of the study, the author applied the EXO method to predict the weather. The result obtained from this study is that the dry season is expected to occur in mid-April 2019 according to the pattern of precipitation data of the land of Husein Sastranegara Airport.