[Paper Review] Forecasting model based on information-granulated GA-SVR and ARIMA for producer price index

Xiangyan Tang, Liang Wang+2 more|arXiv (Cornell University)|Mar 28, 2019

Stock Market Forecasting MethodsDecision Sciences22 references3 citations

TL;DR

This paper proposes a novel hybrid forecasting model that integrates fuzzy information granulation, genetic algorithm-optimized support vector regression (GA-SVR), and ARIMA to improve Producer Price Index (PPI) prediction accuracy. By pre-classifying PPI data into fuzzy information granules, training separate GA-SVR models on each granule, and refining residuals via ARIMA, the model achieves superior accuracy over ARIMA, GRNN, and standalone GA-SVR models in empirical tests.

ABSTRACT

The accuracy of predicting the Producer Price Index (PPI) plays an indispensable role in government economic work. However, it is difficult to forecast the PPI. In our research, we first propose an unprecedented hybrid model based on fuzzy information granulation that integrates the GA-SVR and ARIMA (Autoregressive Integrated Moving Average Model) models. The fuzzy-information-granulation-based GA-SVR-ARIMA hybrid model is intended to deal with the problem of imprecision in PPI estimation. The proposed model adopts the fuzzy information-granulation algorithm to pre-classification-process monthly training samples of the PPI, and produced three different sequences of fuzzy information granules, whose Support Vector Regression (SVR) machine forecast models were separately established for their Genetic Algorithm (GA) optimization parameters. Finally, the residual errors of the GA-SVR model were rectified through ARIMA modeling, and the PPI estimate was reached. Research shows that the PPI value predicted by this hybrid model is more accurate than that predicted by other models, including ARIMA, GRNN, and GA-SVR, following several comparative experiments. Research also indicates the precision and validation of the PPI prediction of the hybrid model and demonstrates that the model has consistent ability to leverage the forecasting advantage of GA-SVR in non-linear space and of ARIMA in linear space.

Motivation & Objective

• To address the challenge of low accuracy in Producer Price Index (PPI) forecasting due to data imprecision and nonlinearity.
• To develop a hybrid forecasting model that leverages the strengths of GA-SVR in nonlinear spaces and ARIMA in linear spaces.
• To enhance prediction precision by applying fuzzy information granulation to preprocess PPI training data into distinct granular sequences.
• To reduce forecasting errors through residual correction using ARIMA modeling on GA-SVR predictions.
• To validate the proposed model’s consistency and superiority over existing models like ARIMA, GRNN, and GA-SVR in PPI forecasting.

Proposed method

• Fuzzy information granulation is applied to pre-classify monthly PPI training data into three distinct sequences of fuzzy information granules.
• Support Vector Regression (SVR) models are individually trained on each granule sequence, with hyperparameters optimized using a Genetic Algorithm (GA).
• The GA-SVR models generate initial PPI forecasts for each granule, producing multiple prediction streams.
• Residual errors between the GA-SVR forecasts and actual PPI values are extracted and modeled using ARIMA to correct systematic forecasting deviations.
• The final PPI forecast is obtained by aggregating the corrected GA-SVR predictions across all granules.
• The hybrid model integrates nonlinear modeling (GA-SVR) and linear error correction (ARIMA) to enhance overall forecasting robustness.

Experimental results

Research questions

• RQ1 Can fuzzy information granulation improve the accuracy of PPI forecasting by reducing data imprecision?
• RQ2 Does combining GA-SVR and ARIMA in a hybrid framework yield better PPI forecasts than standalone models?
• RQ3 To what extent does residual error correction via ARIMA enhance the performance of GA-SVR in PPI prediction?
• RQ4 How does the proposed model compare in accuracy to ARIMA, GRNN, and GA-SVR across multiple evaluation metrics?
• RQ5 Does the hybrid model consistently leverage the complementary strengths of nonlinear (GA-SVR) and linear (ARIMA) modeling in PPI forecasting?

Key findings

• The proposed GA-SVR-ARIMA hybrid model achieves higher forecasting accuracy than ARIMA, GRNN, and standalone GA-SVR models in comparative experiments.
• The integration of fuzzy information granulation significantly improves the model’s ability to handle data imprecision in PPI time series.
• Residual error correction using ARIMA effectively reduces systematic forecasting errors from the GA-SVR component.
• The model demonstrates consistent performance across different data granules, indicating robustness in handling diverse PPI patterns.
• The hybrid approach successfully combines the nonlinear modeling strength of GA-SVR with the linear error correction capability of ARIMA, enhancing overall predictive power.
• Empirical results confirm the model’s validity and superior precision in forecasting the Producer Price Index.