Effect of Kernel in Support Vector Regression Method to Predict Surabaya Consumer Price Index Trend

Staple Food is a necessity that is needed by the people of Indonesia. This is very sensitive when there is an increase or decrease in the price of basic foodstuffs. In the calculations at the cooperative and trade industry offices, some constants are used as benchmarks to assess the increase and decrease in prices of basic foodstuffs, namely the CPI (Consumer Price Index).Support vector regression (SVR) is one of the appropriate methods to solve regression problems based on the concept of maximizing the hyperplane to obtain data that is a support vector. The selection of this method is based on the level of MSE (mean square error) which is lower than the neural network method. As a test material, a sample dataset of staple food prices for the past 5 years was used. Where for training data using a sample of 2013-2014 with a total of 24 records, and testing data using a sample of 2015-2016 as many as 24 records.In this study, the researcher tried to apply multiple kernels in the Support Vector Regression method to predict the CPI (Consumer Price Index). The result of this research is like this.SPLine and Polynomial have less MSE than other kernels it means These two kernels have good usage when use in time-series data.