1 [email protected] [email protected]
Transcription
1 [email protected] [email protected]
PCA [email protected] [email protected] 1 Modeling Forecasting Inflation by Artificial Neural Network Combination model with PCA Abstract: This paper concerning researches are done about reason of deflation in Iran suggest forecasting model for inflation and compare it with ARIMA. Inflation is into annual and use of CPI index. Recently, there has been considerable interest in applications of neural networks in the economics Literature but In contrast, relatively few studies have applied neural network methods to macroeconomic time series. So we try forecast this macroeconomic variable precisely by enter factor that reflate. Variable that derive of previous studies are: liquidity, exchange rate, interest rate, Imported good and previous inflation for Expected Rate of Inflation. Conclusions indicate that neural network forecast precisely and entrance effective variables on inflation improve accurate forecasting. Keyword: forecasting/inflation/artificial neural network/liquidity/interest rate/imported goods index/ Expected Rate of Inflation/PCA 2 ARIMA CPI ARIMA MSE AR(1) C45,C53,E31,E37 JEL / 3 [1] [12,13] [14,15] ARIMA STRS NNRS NNRS [16] CPI ARIMA 1 2 James nakamura Neural Network Regime Switching 4 Smoot-Transition Regime Switching 5 In sample performance 6 Out-of-Sample performance 7 Customer price index 3 4 [2] [3] [4] 5 [5] [17] ARIMA artificial neural network Mackey-Glass Equation Methods Threshold AR Model Time-Varying Parameter 1 2 3 Expected Rate of Inflation threshold autoregressive time-varying 6 [9] [10] f b W . P a feed forward 1 Classify Cluster 3 feed forward 4 recurrent 2 7 [6] a MSE w t [-1,1] pn = 2*(p-minp)/(maxp-minp) – 1 p pn [18] [-1,1] [0,1] [19] 1 Back Propagation Imported good 3 normalization 4 Component principle analysis 2 8 feed forward tansig MSE,RMSE,MAE,MAPE ARIMA AR(1) Q -5.50626 -3.73785 -2.99188 -2.63554 ARIMA . 1 Kolmogorov Theorem Unit Root Test 3 Auto Correlation 4 Partial Auto Correlation 2 9 2 p (d p zp) p 1 1 MSE ( MeanSquaredError ) p 2 p (d p zp) p 1 2 RMSE ( RootMeanSquaredError ) p 2 p (d p 2 3 R (CofficientOfDeter min ation) 1 zp) p 1 2 p (d p dp) p 1 p dp 4 MAE ( MeanAbsoluteError ) zp p 1 p 5 MAPE ( MeanAbsolutePercentageError ) 100 P p p 1 dp zp dp R AR(1) . 10 MSE RMSE MAE MAPE AR(1) ARIMA ARIMA 60 50 40 30 20 10 0 1 spss13 3 5 7 9 11 EWiews 13 15 17 19 21 23 25 27 29 31 33 35 AR(1) MATLAB7 Minitab13 MSE AR(1) 11 – 9- kate A.Smith, jatinder N.D. Gupta-Hersheg.(2002).neural network in Business: techniques and application. PD.Idea Group pub. 10-Lakshminarayanan sriram.(2005). an integrated stock market forecasting model using neural network. College of engineering and technology of ohio university.pp 1-23. - James H. Stock , Mark W. Watson.(1999). Forecasting inflation. Journal of Monetary Economics.pp 293-335. -Peter Macadam , Paul McNeil’s.(2005). Forecasting inflation with thick models and Neural networks. Economic Modeling 22.pp 848-867. - Cláudia Duarte, António Rua.(2007). Forecasting inflation through a bottom-up approach: 1 How bottom is bottom?.Economic Modelling.pp1-13. - Emi Nakamura.(2005). Inflation forecasting using a neural network. Economics Letters 1 86.pp 373-378 16- Paul D. M C Nelis-Amsterdam.(2005).neural network in finance :Gaining predictive Edge in the market . Elsevier Academic press.pp167-197. 17- Bina R. Setyawati(2005),Multi lyer feed forward neural networks for forien exchange time series forecasting.Department of Industrial and Management Systems Engineering Morgantown, West Virginia.pp 1-61. 18-Jang JSR,(1993) ANFIS: Adaptive-Network-based Fuzzy Inference System.IEEE Transaction on System,Man, and Cybernetic;665-685. 19-Kim Yong Seog,(2004),An intelligent system for targeting:a data mining approach.J Decision Support System;215-228. 12 VAR00001 27 N Normal Parameters(a,b) Most Extreme Differences Mean 1.5214 Std. Deviation 6.58598 Absolute .132 Positive .132 Negative -.097 Kolmogorov-Smirnov Z .685 Asymp. Sig. (2-tailed) .737 a Test distribution is Normal. b Calculated from data. AR(1) VAR00001 N Normal Parameters(a,b) Most Extreme Differences 27 Mean 2.2130 Std. Deviation 9.65110 Absolute .104 Positive .072 Negative -.104 Kolmogorov-Smirnov Z .539 Asymp. Sig. (2-tailed) .933 a Test distribution is Normal. b Calculated from data 13 1350 9.091 296.30 76.0 9.5 9.5 1.1 1351 9.09091 399.4 69.0 9.5 9.5 1.2 1352 8.33333 517.5 68.0 9.5 9.5 1.3 1353 15.3846 813.7 68.0 9.5 9.5 1.5 1354 6.66667 1149.5 68.0 9.5 9.5 1.6 1355 18.75 1625.7 71.0 10.5 10 1.7 1356 26.3158 2139.4 71.0 10.5 10.5 1.9 1357 8.33333 2578.6 100.0 10.5 10.5 2.1 1358 11.5385 3628.3 141.0 6 8 2.4 1359 24.1379 4508.1 200.0 6 8 2.9 1360 1361 22.2222 20.4545 5236.1 6430.7 270.0 350.0 6 6 8 8 3.2 3.5 1362 1363 1364 13.2075 11.6667 5.97015 7514.4 7966.9 9002.1 450.0 580.0 614.0 6 10 10 8 8 8 3.7 3.8 4.1 1365 23.9437 10722.6 742.0 10 8 5.3 1366 28.4091 12668.2 991.0 10 8 7.1 1367 28.3186 15687.6 1019.0 10 8 8.8 1368 17.931 18753.3 1207.0 10 8 10 1369 9 22969.5 1412.0 14 12 13.4 1370 20.7 28628.4 1420.0 14 12 16.2 1371 24.4 35866.0 1498.0 14 13 22.5 1372 22.9 48135.0 1806.0 14 17 28.7 1373 35.2 61843.9 2667.0 15 17 42.2 1374 49.4 85072.2 4036.0 14.5 18 72.5 1375 23.2 116552.6 4446.0 14.5 18 93.6 1376 17.3 134286.3 4782.0 14.5 18 100 1377 18.1 160401.5 6468.0 15.5 18 110 1378 20.1 192689.2 8658.0 15.5 18 134.2 1379 12.6 249110.7 8188.0 15.5 18 152.1 1380 11.4 320957.3 8008.0 15.5 17 153.3 1381 1382 1383 15.8 15.6311 15.2393 417524.0 526596.4 685697.5 8019.0 8323.0 8747.0 14.5 15 15 16 16 15 159.7 167.3 191.5 1384 12.0583 921019.4 9042.0 15 16 204.2 1385 15.3 1284199.4 9244.0 13 14 226.2536 14