A Novel Machine Learning-based Strategy for Agricultural Time Series Analyzing and Forecasting: a Case Study in China's Table Grape Price

Abstract

Applications of data science for agriculture has been widely discussed, this study attempts to construct a novel machine learning-based strategy for products price analyzing and forecasting. To do this, we follow the framework of "divide and conquer" to strategically integrate the Ensemble Empirical Mode Decomposition (EEMD), reconstruction algorithms, evolutionary Least Squares Support Vector Machine (LSSVM) and Extreme Learning Machine (ELM) to realize numerical forecasting and qualitative analysis. In the price prediction scenario of table grape, which is a typical perishable fruit in China's fruit market, the performance of the proposed method is verified. This paper is committed to provide a reference for the univariate time series price analysis of perishable agricultural products when the conditions are not enough to analyze the influencing factors, free it from the tedious process of data collection, and realize the accurate prediction and qualitative analysis of the target series.