Weather Classification and Prediction on Imagery Using Boltzmann Machine

Abstract

Weather is a physical process or event that occurs in the atmosphere at a specific time and place, as well as its changes over a short period in a particular location on Earth. To produce weather forecast information, there is a series of processes that must be carried out until the weather information is conveyed accurately. The stages involved in the feature extraction process are carried out first. This process is carried out to obtain specific characteristics or features from a dataset. After the feature extraction process has been completed, the next step is to predict the weather based on the input images. To classify the weather on Earth using various algorithms, one of which is the Machine Boltzmann. The pattern recognition method used is Machine Boltzmann as an application of a simpler and more complex method. Generally, the weather prediction system using Machine Boltzmann consists of several stages, namely image acquisition, greyscale processing, segmentation/location using Sobel edge detection, classification using the Machine Boltzmann method, and finally producing output in the form of weather class results. The classification process in this research involves images of clear, cloudy, and rainy weather as inputs. The output of the system is the determination of whether the input weather image falls into the category of clear, cloudy, or rainy weather. The results of the study show that the classification of weather based on captured images has the highest accuracy for clear weather, with a percentage of 73.33%. For cloudy weather, the success rate is equal to the error rate, which is 50%, while rainy weather was not recognized at all.