Ensemble Prediction of China Consumer Price Index Based on MLP and BiLSTM

DOI: https://doi.org/10.62517/jbdc.202501414

Author(s)

Zhuoheng Song, Cong Gu*, Canhui Zhang, Shuo Zhao

Affiliation(s)

School of Mathematics and Information Science, Zhongyuan University of Technology, Zhengzhou, Henan, China *Corresponding Author

Abstract

Accurate prediction of the Consumer Price Index (CPI) is crucial for policymakers to grasp a country’s economic operation rules and has a significant impact on policymaking and resource allocation. However, the nonlinear and nonstationary characteristics of financial data pose significant challenges to achieving accurate and robust predictions. This paper proposes a novel multi-step forecasting ensemble model that integrates Hiking Optimization Algorithm-optimized Variational Mode Decomposition (VMD), Multilayer Perceptron (MLP), and Bidirectional Long Short-Term Memory (BiLSTM) networks called HOA-VMDMLP-BiLSTM (dynamic weighting). Initially, cubic spline interpolation transforms monthly data into weekly frequency to address sample size limitations. Subsequently, the Hiking Optimization Algorithm (HOA) adaptively determines the optimal decomposition modes in VMD, obtaining the independent and stationary components while reducing noise interference. The BiLSTM and MLP networks then respectively process the decomposed modes and interpolated weekly series, with their predictions dynamically weighted through inverse error variance weighting to get the final value. Experimental results show that the Model’s determination coefficients (R²) values for 1 step, 5 step, and 9 step predictions are 0.9964, 0.9776, and 0.9411 respectively. The Diebold-Mariano test rejects the null hypothesis at the 1% significance level, indicating the proposed model’s statistical superiority over benchmark methods. Notably, the proposed model demonstrates not only excellent in 1 step prediction but also robust and reliable in multi-step forecasting.

Keywords

CPI Forecast; HOA-VMD; MLP; BiLSTM; Ensemble Prediction

References

[1] Zhang, Y. Evaluation of Influencing Factors of National Engineering Project Cooperation along the ‘Belt and Road’ Based on Entropy Weight TOPSIS. Science Technology and Industry, 2025, 25(05): 216-223. [2] Alvarez-Diaz, M. and Gupta, R. Forecasting US Consumer Price Index: Does Nonlinearity Matter. Applied Economics, 2016, 48(46): 4462-4475. [3] Sun, H. Comparative Analysis of China’s CPI Dynamics under Two Financial Crises. Statistics & Decision, 2010, 26(1): 116-118. [4] Lu, S., Zhao, B., Bi, J. CPI Forecasting Integrating Fuzzy Information Granulation and SVM Time Series Regression. Statistics & Decision, 2015, 14: 82-84. [5] Zeng, L. CPI Prediction Based on ARIMA-RF Combined Model. Modern Information Technology, 2023, 7(13):13-17. [6] Sarangi, P.K., Sahoo, A.K., Sinha, S. Modeling Consumer Price Index: A Machine Learning Approach. Macromolecular symposia, 2022, 40(1): 2100349. [7] Wen, X. and Xu, Y. EMD-based Analysis and Predictive Modeling of Wind Turbine Power Characteristics. Acta Energiae Solaris Sinica, 2021, 42(11): 293-298. [8] Tai, X. and Liu, Y. Monthly CPI Forecasting using a Hybrid EEMD-PSO-SVM Model. Statistics & Decision, 2019, 3: 30-33. [9] Ying, H. and Wang, S. CPI Forecasting using a Hybrid EEMD-BP Model. Contemporary Economics, 2022, 39(10): 97-106. [10] Fang, Z., Yang, H., Deng, Q., Fang, H., Zhuang, Y., Lin, S., Yao, W. Dissolved Oxygen Prediction Research based on Correlation Optimization with the Coupled LSTM Model. Safety and Environmental Engineering, 2025: 1-17. [11] Zhou, Y., Peng, J., Bai, Y. Data-driven Time Series Feature Extraction and Prediction of Daily Blood Collection Volumes. Journal of Systems Science and Mathematical Sciences, 2025: 1-22. [12] Dong, M. and Tang, X. Study on CPI Prediction by LSTM Model based on Double-layer Attention Mechanism. Chinese Journal of Management Science, 2025: 1-14. [13] Dragomiretskiy, K. and Zosso, D. Variational Mode Decomposition. IEEE Transactions on Signal Processing, 2013, 62(3): 531-544. [14] Zhang, Y., Zhao, Y., Kong, C., Chen, B. A New Prediction Method based on VMD-PRBF-ARMA-E Model Considering Wind Speed Characteristic. Energy Conversion and Management, 2020, 203: 112254. [15] Feng, Y., Zeng, H., Deng, H., Tu, P. A Step-type Landslide Displacement Prediction Model based on Creep Trend Influence and Feature Optimization Algorithm. Chinese Journal of Rock Mechanics and Engineering, 2025, 44(03): 705-720. [16] Zhou, W., Li, J., Cai, Y., Zheng, Z., Wen, L. Step-like Landslide Displacement Prediction using Bovmd-p-boxgboost. South-to-North Water Transfers and Water Science & Technology, 2025: 1-19. [17] Oladejo, S.O., Ekwe, S.O., Mirjalili, S. The Hiking Optimization Algorithm: A Novel Human-based Metaheuristic Approach. Knowledge-Based Systems, 2024, 296: 111880. [18] Goodchild, M.F. (2020). Beyond tobler’s hiking function. Geographical Analysis, 52(4), 558-569. [19] Li, H., Xiao, B., Jin, K., Chen, Z., Yu, L. Construction and Comparative Analysis of Water Quality Prediction Model of the Sanmenxia Reservoir of the Yellow River. Environmental Engineering, 2024, 42(12): 1-7. [20] Zhang, F., Zhang, Q., Qin, B., Zhu, G., Deng, X., Liu, D. Bearing Life Prediction based on VMD and TCN-SeNet-BiLSTM Networks. Machine Tool & Hydraulics, 2025, 53(1): 15-23. [21] Zhang, J., He, J., Cai, Q., Kang, Z., Yang, Y., Wang, T., Yang, L. Prediction of Cod Concentration in Wastewater Treatment Plant Effluent based on Secondary Decomposition and BiLSTM. CIESC Journal, 2025: 1-16.