WEATHER PREDICTION MODEL BASED ON METAR DATA AND SYNOPTIC OBSERVATION STATION USING DATA MINING METHODS

• Main Author: Anggun Novembra, Muhammad
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/36570
• Institution: Institut Teknologi Bandung

This Research discusses prediction models of several phenomena using data mining models. The algorithm that will be used is the kNN classification algorithm, Naïve Bayes, ANN and SVM. Previous research has proven that data mining models use trials that are very good at predicting the phenomenon of rain or rain. This study tries to predict more than a few studies using one model. The knowledge extraction from historical weather data collected from the site rp5.ru. There are two types of weather report formats, the first dataset is Weather Synoptic Station with 29 attributes, the second dataset is METAR with 13 attributes. The condition of each dataset is inconsistent, so this study uses feature reduction and feature selection using the information gain method. In addition, simplification of the class is done to handle the distribution of uneven weather event classes. In the synoptic dataset there are 47 weather phenomena and in the METAR dataset there are 25 weather phenomena, simplified into 8 weather phenomena in the synoptic dataset and 5 weather phenomena on the metar dataset based on the similarity of weather phenomena. This study uses FScore as a measurement variable. Based on the experimental results from the data mining model using the kNN algorithm it produces better Fscore than experiments with using the other algorithms. The results of the Fscore METAR dataset experiment are 76.9% using four features : relative humidity (U), total cloud cover (c), atmospheric pressure at station level (Po), and horizontal visibility (VV). The results of the Fscore weather station dataset experiment synoptic observations are 75.4% using ten features : average wind speed (ff), total cloud cover (N), relative humidity (U), atmospheric pressure at the level station (Po), atmospheric pressure on sea level (P), wind direction (DD), weather conditions before observation 1 and 2 (W1 and W2), cumulonimbus, cumulus, stratus, stratocumulus (Cl) cloud conditions.