Abstract:
Pekanbaru City is a city with a tropical climate with a minimum air temperature of20.2ºC-23.0ºC and
a maximum of 34.1ºC-35.6ºC and precipitation of 38.6- 435.0mm/year with the average rainy season
from September until April. A prediction of precipitation is required to avoid the impact caused by
the high precipitation. This research aims to predict the precipitation period of Pekanbaru City from
2016 until 2020 using the Back Propagation Neural Network method. The data evaluated in monthly
data such as precipitation, wind speed, air temperature, humidity, air pressure, and duration of
sunlight. This method uses the Matlab R2018a program for programming data and Microsoft Excel
for evaluating data and grouping data respectively. The architecture used is composed of 5 input
layers using the logsig activation function, 5 hidden layers using the logsig activation function, and 1
output layer using the purelin function. This study divides the data into two parts, i.e., from 2016
until 2019 as training data and from 2020 as a test. The resultsfor the CGB train network training
produced the smallest MSE value of 0.00015336 and the network testing produced the best
precipitation prediction in February with an error of 0.99736598% where the precipitation was
163.63201mm and the precipitation data was BMKG 162mm.
