基于各模型对重庆GDP实证分析

Empirical Analysis of Chongqing GDP Based on Various Models

摘要: 分析1979年至2021年的重庆GDP数据的时间序列，建立了自回归滑动平均求和模型ARIMA(p,d,q)，并用该模型预测的2020年和2021年重庆GDP数据与实际数据进行比较，对建立的模型进行优化评估，最后利用优化模型对2022年和2023年重庆GDP进行短期预测，为重庆经济的发展提供参考。根据建立的时间序列分析得到最优模型为ARIMA(0,2,2)，预测值与实际值的平均绝对误差为1.22%，能较好地反映重庆GDP发展的趋势并进行短期预测。此外还对语言自动识别的ARIMA(0,2,1)模型进行分析，并基于BP神经网络、CNN、LSTM模型等对重庆GDP进行预测。在2020年中，基于ARIMA(0,2,2)模型预测结果的相对误差的绝对值最小。在2021年中，基于单变量MLP模型预测结果的相对误差的绝对值最小，在平均绝对误差水平中，ARIMA(0,2,2)模型的平均绝对误差水平最小。综合可以看出ARIMA(0,2,2)模型的预测效果是最好的。

Abstract: This paper analyzes the time series of Chongqing’s GDP data from 1979 to 2021, establishes an autoregressive moving average summation model ARIMA(p,d,q), compares the GDP data predicted by this model in 2020 and 2021 with the actual data, optimizes and evaluates the established model, and finally uses the optimized model to make a short-term prediction of Chongqing’s GDP in 2022 and 2023, providing reference for Chongqing’s economic development. According to the established time series analysis, the optimal model is ARIMA(0,2,2), and the average absolute error between the predicted value and the actual value is 1.22%, which can better reflect the development trend of Chongqing’s GDP and make short-term prediction. In addition, the ARIMA(0,2,1) model of automatic language recognition is analyzed, and the GDP of Chongqing is predicted based on BP neural network, CNN and LSTM model. In 2020, the absolute value of the relative error of the results predicted based on ARIMA(0,2,2) model is the smallest. In 2021, the absolute value of the relative error of the results based on the univariate MLP model is the smallest, and the average absolute error level of ARIMA(0,2,2) model is the smallest. It can be seen that ARIMA(0,2,2e) modl has the best prediction effect.