ALGORITMA CUCKOO SEARCH DAN BACKPROPAGATION LEVENBERG MARQUARDT UNTUK MEMPREDIKSI CUACA DI KOTA PALEMBANG
Daftar Isi:
- Weather prediction is an activity related to the application of science and technology to estimate future atmospheric conditions. In this research, BPLM (backpropagation Levenberg Marquarts) artificial neural training algorithm is used to build a weather prediction model, but this method have shortcoming where the training process is highly dependent on initial random weights before training stage begin. Therefore, the CS optimization algorithm (cuckoo search) is used to initialize the initial weight before training using BPLM. Weather prediction is done by dividing the prediction class 5 categories, namely very mild, mild, medium, heavy and very heavy rain. The data used is sourced from BMKG and is a daily weather element data in the city of Palembang. The test results show that both methods CSLM (cuckoo search backpropagation Levenberg Marquarts) and BPLM is only able to predict 3 out of 5 weather categories very mild, mild and medium rain. CSLM method produces average validation accuracy 59.58% and average testing accuracy 63.63%.