Yield Prediction of Household Garbage Based on SARIMA and Exponential Smoothing Model

Abstract

Refuse classification is a waste treatment method, which creates reuse opportunities and enhances recyclable value for domestic waste. Prolonging the use cycle of renewable resources is one of the cores of China's sustainable development strategy. Analyzing the total amount of household waste can obtain the characteristics of regional garbage growth so that governments can promote the garbage classification process intelligently. This article chose the household garbage monthly yield data as the time series and extracted its seasonal information. We used RStudio software to develop the multiplicative seasonal autoregressive integrated moving average (SARIMA) model with 2 Holt-Winters exponential smoothing models (additive and multiplicative) for comparison. Seasonal factor charts explain that the seasonal effect cycle of series is 1 year, and its period of increase or decrease is 6 months. SARIMA(0,1,1)(1,1,0)12 is the optimal model for data, which is better than Holt-Winters models. The prediction accuracy of SARIMA model is 94.31%. For the next year, the SARIMA model predictions show that the growth rate of the household garbage yield is 2.65%, which is in line with the actual situation. The results of this study are valuable for the planning of waste classification policy in recent years.

A Time Series Data Regression Model Based on Multi-kernel KPCA Dimension Reduction and Bagging Algorithms

“…Multiple regression analysis is a statistical method to establish a forecasting formula (empirical formula) based on statistical data [2]. For example, Zhang Jie [3] et al predicted the total amount of garbage based on multiple regression Li Jinhai et al [4] put forward the application steps of the multiple regression method based on the multiple regression mathematical model. Wang Quanzhong et al [5] introduced two types of multiple regression models, but the multiple regression models are prone to fall into curse of Dimension when the predictor is high dimension.…”