基于互联网的农产品价格空间传导分析研究

发布时间：2017-12-28 16:12

本文关键词：基于互联网的农产品价格空间传导分析研究　出处：《中国科学技术大学》2017年硕士论文　论文类型：学位论文

【摘要】：我国农产品买难卖难问题日益凸显,严重影响农民收益和农业经济秩序稳定。农产品价格传导作为价格机制中不可或缺的组成内容,对于及时发现商品交易过程中买难卖难现象具有重要意义。而目前的农产品价格传导研究主要利用国家、省份的农业相关统计年鉴数据,存在严重滞后且数据粒度较粗,无法及时发现商品价格在不同城市间传导规律。随着信息化技术的快速发展,互联网已经成为最重要的信息源头,网络上农产品电商平台超过30000家,每天定时发布更新农产品公告、供求、价格等消息,数据实时且价格可细化到各个城市农贸市场一级,因此利用这些平台数据就有可能更加精确发现农产品价格空间传导路径。本文重点以互联网大数据为背景,针对农产品价格在不同地域的传导方向、周期等内容开展研究,具体来说,包括以下3部分。(1)由于互联网信息缺乏统一的描述格式,数据冗余且表达不一致,因此首先对农产品名称与类别进行语义级的自动识别。而传统的基于条件随机场(CRF)的信息抽取方法在农产品名称与类别识别过程中依赖样本语料,人工标注工作量大,没有考虑语义特征,准确率低。因此本文为解决该问题,利用CRF结合农业本体的方法进行农产品名称与类别识别。共使用词、词性、地理属性和农业本体概念4种特征,完成4组对比实验,识别7种农产品类别。通过实验表明本体概念的加入使CRF开放测试的总体准确率、召回率、F值提高了 10.20%、59.78%以及35.17%,从而有效实现了互联网农产品相关信息的语义级获取。(2)以西红柿为例,选取北京、上海、重庆等26个城市的从2016年1月11日到2017年2月27日的以3天为周期的共计138条价格时间序列数据。以300千米为半径将城市划分为8个区域,分别对同一区域内的城市西红柿价格数据进行平稳性判断即ADF单位根检验、长期稳定性衡量即Johansen协整分析并实现价格传导方向判断的Granger因果检验。综合考虑空间与时间因素,估算出价格空间传导的合理周期并得到传导路径,有利于减缓价格大幅度波动。(3)基于西红柿价格空间传导路径研究整理出来的农产品相关数据,开发了可视化系统,可实现以时间、价格或地区等不同维度的可视化展示,提高了决策的可交互性。
[Abstract]:The problem of difficult to buy agricultural products is becoming more and more difficult in China, which seriously affects the farmers' income and the stability of the agricultural economic order. As an indispensable component of the price mechanism, the price transmission of agricultural products is of great significance to the timely discovery of difficult buying and selling in the process of commodity trading. However, the current agricultural price transmission research is mainly based on the agricultural related statistical yearbook data of the country and the provinces, which is seriously lagged behind and the data granularity is relatively coarse, which can not find the transmission rule between commodity prices in different cities in time. With the rapid development of information technology, the Internet has become the most important source of information on the network, agricultural business platform more than 30000, daily update of agricultural products supply and demand, price announcements and news, real-time data and the price can be refined to each city farmers market level, so the use of these platform data may be more accurate the price discovery space transmission path of agricultural products. Based on the Internet big data, this paper focuses on the direction and cycle of agricultural products prices in different regions. Specifically, it includes the following 3 parts. (1) due to the lack of unified description format of Internet information, redundant data and inconsistent expression, we first automatically identify the semantic level of agricultural product names and categories. The traditional information extraction method based on conditional random fields (CRF) relies on sample corpus in the process of identifying agricultural product names and categories, and the workload of manual annotation is large without considering the semantic features and low accuracy. Therefore, in order to solve this problem, the name and category of agricultural products are identified by CRF combined with the method of agricultural ontology. A total of 4 types of agricultural products are identified by 4 groups of comparative experiments, including the common use of words, words, geographical attributes and agricultural noumenon concepts. Experiments show that the overall accuracy, recall rate and F value of CRF open test increased by 10.20%, 59.78% and 35.17%, which effectively realized the semantic level acquisition of Internet related agricultural products information. (2) taking tomatoes as an example, 138 price time series data with 3 days from January 11, 2016 to February 27, 2017 were selected from 26 cities in Beijing, Shanghai and Chongqing. The 300 km radius of the city will be divided into 8 regions, respectively, on the same area of the city of tomato price data to determine stability of ADF unit root test, measure the long-term stability of Johansen cointegration analysis and Granger causality test to determine the direction of price conduction. Considering the factors of space and time, the reasonable cycle of price space conduction is estimated and the conduction path is obtained, which is beneficial to slow down the large fluctuation of price. (3) based on Tomato's spatial transmission path, we have developed a visualization system, which can realize visual display in different dimensions such as time, price or region, and improve the interactivity of decision-making.
【学位授予单位】：中国科学技术大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F323.7;TP391.1

【相似文献】