Predicting GDP of Indonesia Using K-Nearest Neighbour Regression

Bagus Priambodo,Asama Kudr Nseaf,Sarwati Rahayu,Mardhiah Masril,Zico Pratama Putra,Inge Handriani,Emil Naf’An,Al Hamidy Hazidar,Yuwan Jumaryadi,Deni Setiawan

doi:10.1088/1742-6596/1339/1/012040

Abstract

The impact of the global recession in 1998 that originated from the recession in the US will affect the projected economies in Asia, including Indonesia, both direct and indirect nature. In this study, we predicted Indonesia’s GDP in the event of the economic crisis that hit Indonesia starting in 1998. Instead of using the famous prediction algorithm as a neural network and linear regression. K-Nearest Neighbour is selected because it is easy and fast to use in the small dataset. We use a dataset from 1980-2002, consisting of rice prices, premium prices, GDP of Japanese country, American GDP, currency exchange rates, Indonesian government consumption, and the value of Indonesia’s oil exports. For evaluation, we compare k-NN regression prediction result with prediction result using back propagation neural network and multiple linear regression algorithm. Result show, k-NN regression is able to predict Indonesia’s GDP using small dataset better than the neural network, and multiple linear regression method.

Highlights

Related Work and ProblemMany studies have applied K-NN to solve the problem in many fields for classification and regression [8]
Bagus Priambodo1*, Sarwati Rahayu2, Al Hamidy Hazidar3, Emil Naf'an4, Mardhiah masril5, Inge Handriani6, Zico Pratama Putra7, Asama Kudr Nseaf8, Deni Setiawan9, Yuwan Jumaryadi10
The understanding of the phenomenon of Indonesia's economy and its relationship with the possibilities that occur in the future are very helpful in determining attitudes or policies that lead to hope or expectation to be achieved

Summary

Related Work and Problem

Many studies have applied K-NN to solve the problem in many fields for classification and regression [8]. Some studies use k-NN for classification, predict heart disease [9] K-NN prediction result show better accuracy compare with naïve Bayes and decision tree, classification of the poor for health cards distribution[10] the accuracy of the results of determining feasibility using a combination of K- Nearest. In economic field many studies have been done in predicting GDP and inflation, using various method and algorithm, most of them use the neural network and linear regression [3], [4], [11], [13], [14], there was study in predicting GDP using k-NN. We will evaluate whether the K-NN prediction results are better than using neural networks and multiple linear regression for short time prediction

Methods

Multivariate prediction

Result and Discussion

Conclusions