Pattern Prediction Comparison of Time Series Data Using Artificial Neural Network (ANN) - Multilayer Perceptron (MLP) and Support Vector Regression (SVR)

Abstract
Authors
Keywords
Conclusion
References

Support Vector Regression (SVR) and Artificial Neural Network (ANN) - Multilayer Perceptron method has been proven to be able to solve time series forecasting problems to achieve generalization performance. Both methods are also able to handle small and large sample data. In this study, these two methods will be compared to find the most accurate. The data used is time series data in the form of secondary data about licensing, the season before delegation of authority. The data will be trained and tested using the Python programming language using the SVR method with several kernel functions and MLP methods. The composition of the use of training data and testing data is 70% and 30%. Then with these data, a Simple Linear Regression calculation will be conducted to test the extent to which the causal relationship between the number of permits issued and the number of available officers, so that the effectiveness of licensing service performance can be known. The results showed that with small sample data, the accuracy of the MLP method proved to be able to do better forecasting than the SVR method with several kernel functions with an accuracy level of MLP score -0.17 and the values of MSE, MAE and RMSE are 251.09, 11.45, and 15.84. Based upon SLR calculations, it is evident that there is a significant relationship between the variables x and y, namely the number of service personnel influencing the effectiveness of service performance and the number of permits that have been processed.

Published In : IJCAT Journal Volume 6, Issue 6

Date of Publication : June 2019

Pages : 37-45

Figures :09

Tables :02

Ika Oktavianti : Department of Computer Engineering Faculty of Computer Science, Sriwijaya University, Indonesia.

Ermatita : Department of Computer Engineering Faculty of Computer Science, Sriwijaya University, Indonesia.

Dian Palupi Rini : Department of Computer Engineering Faculty of Computer Science, Sriwijaya University, Indonesia.

Pattern prediction; time series data; multilayer perceptron; support vector regression

The main purpose of this study was to compare the performances of the Perceptron-Multilayer Neural Network (MLP-ANN) and Support Vector Regression (SVR) to predict licensing patterns and forecast the number of licenses to be issued in the following month and year. This study used data series based on monthly data from August 2009 to January 2016. In addition, we used various kind of kernels to improve performance of SVR model. That is Linear, Linear Model Lasso, Elastic, Ridge and RBF kernel. The results showed the ability of using time series data which proves that the MLP-ANN method is more accurate than the SVR method, where the accuracy value is -0.17 and the Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) values are 11.45 and 15, 84. The major difficulty of the work was to achieve minimize generalized error bound so that the difference of actual value and the predicted value is close to zero. Based upon SLR calculations, it is evident that there is a significant relationship between the variables x and y, namely the number of service personnel influencing the effectiveness of service performance and the number of permits that have been processed. As a future work, we recommend using hybrid approaches, which are the combination of ANN and SVR. Otherwise it is expected to select the most appropriate parameters of the kernel function and ANN model. Another future improvement can be proposed such as using metaheuristic algorithm, deep learning and genetic algorithm techniques in the forecasting accuracy.

[1] A. Xu and M. Raginsky, "Information-theoretic analysis of generalization capability of learning algorithm," 31st Conference on Neural Information Processing System (NIPS) CA, USA, 2017. [2] M. Khashei, S. R. Hejazi and M. Bijari, "A new hybrid artificial neural network and fuzzy regression model for time series forecasting," Elsevier Fuzzy Set and System, vol. 159, pp. 769-786, 2008. [3] Turban, E. 2005. Decision Support Systems and Intelligent Systems. Edisi Bahasa Indonesia Jilid 1. Andi: Yogyakarta. [4] Tri Wulandari, Retno. 2017. Data Mining Theory and Rapidminer Application. Gava Media: Yogyakarta. [5] Fayyad, Usama. 1996. Advances in Knowledge Discovery and Data Mining. MIT Press. [6] Larose, Daniel T. 2005. Discovering Knowledge in Data: An Introduction to Data Mining. John Willey & Sons, Inc. [7] Kasmir, Jakfar. 2003. Business Feasibility Study. Jakarta: Prenada Media. [8] Cryer, J.D., dan Chan, KS. 2008. Time Series Analysis: With Application in R: Second Edition. USA: Spinger Science dan Business Media, LLC. [9] Santosa, Budi. 2007. Data Mining Techniques for Using Data for Business Purposes. Graha Ilmu: Yogyakarta. [10] D. Basak, S. Pal and D. C. Patranabis, "Support vector regression," Neural Informating Processing. 11, no. 10, 2007. [11] V. Vapnik, S. Golowich and A. Smola, "Support vector method for function approximation, regression estimation, and signal processing," Neural Information Processing Systems, MIT Press Cambridge, vol. 9, 1997. [12] P. Kumar and P. Sharma, "Artificial neural network-a study,"International Journal of Emerging Engineering Research and Technology, vol. 2, issue 2, pp. 143-148, 2014. [13] J. M. Nazzal, I. M. El-Emary and S. A. Najim, "Multilayer perceptron neural network (MLPs) for analyzing the properties of Jordan oil shale," Wolrd Applied Sciences Journal, vol. 5, (5) pp. 546-552, 2008. [14] C.J. Lu., T.S. Lee., C.C. 2009. Chiu. Financial Time Series Forecasting Using Independent Component Analysis and Support Vector Regression. Taiwan. [15] Adiningtyas, Desy Trishardiyanti. 2015. Forecasting the Number of Hotel Guests in Demak Regency Using the Support Vector Regression Method. Diponegoro University, Semarang. [16] Dong, B., Cao, C., Lee, S.E., 2005. Applying Support Vector Machines to Predict Building Energy Consumption in Tropical Region. Energy and Buildings 37, 545-553.