农产品市场价格web信息分析方法研究

发布时间：2018-11-13 10:44

【摘要】：近年来,我国农产品市场价格呈现的异常波动已成为社会关注的焦点。农产品价格波动不仅会对农民收入和农民生产积极性产生直接影响,更关乎百姓的日常生活和切身利益。为保持经济平稳健康发展、保障群众生活,稳定物价的宏观调控尤为重要。价格监测和预测是维持价格稳定的一个重要环节,精确地价格监测和低误预测是涉农科研工作者的一个重要研究方向。我国农产品市场价格Web信息分布广、更新快,迫切需要建立一套垂直搜索引擎系统实现定期抓取网络中农产品价格数据,迫切需要建立一个农产品价格分析、预测和监测平台,提供全面、清晰的分析结果。为政府部门管理人员提供生产调控、决策分析的依据,成为农民种植植物的决策依据,为农产品市场价格稳定做出积极的贡献。某些农网等价格信息网提供的农产品价格信息有数据单位不统一、产品名称不规范等问题,经过对分布在不同网站上的农产品市场价格数据的分析和总结,提出了规范产品名称、规范市场名称、初始化农产品类别、初始化省市、规范数据单位、去重策略选择和零价格数据处理7个数据规范化原则。研究了DOM树方式、正则表达式、HTMLParser提取网页文本信息,利用Heritrix等软件搭建了农产品市场价格垂直搜索引擎系统实现了抽取不同农网上的价格信息,经规范化后形成了统一、完整的SQL server农产品价格数据库。为进一步提高农产品市场价格预测精度,及时发现价格异常的农产品,选取了山西晋城绿欣农产品批发市场胡萝卜、白萝卜、大白菜、大葱、豆角、黄瓜、尖椒、韭菜、茄子、青椒、土豆、西红柿和油菜十三种农产品进行预测算法对比研究。在加权算术平均预测法中对比分析了5种权数设置方法,实验结果表明以当年价格为权数误差最低,优于其它权数设置方法,在平均数预测法中优于简单算术平均法；对比分析了时间序列非季节11种预测方法,研究结果表明二次曲线趋势延续法和龚伯兹曲线趋势延续法不适用于农产品价格预测,在简单算术平均法、加权算术平均法、时间序列平均增长量预测法、时间序列几何平均法、一次移动平均预测法、二次移动平均预测法、一次指数平滑法、二次指数平滑法和直线趋势延续法9种预测方法中二次移动平均预测法和二次指数平滑法误差低于其它预测方法,适用于农产品价格预测,在此基础上提出了一种改进的二次指数平滑预测法,二次指数平滑预测法中一次、二次平滑系数不同时,所有的误差平方和都小于或等于一次和二次平滑系数相同时的误差平方和。改进后的二次指数平滑预测法误差最低,优于末改进的二次指数平滑法,也优于二次移动平均预测法；对比了时间序列季节指数水平法和季节指数趋势法,从实验结果中可以看出,大多数农产品两种预测方法误差平方和相差很多,实际预测价格可以采用误差平方低的预测值；在价格异常农产品判定方面提出了从预测值与实际价格误差平方和历史误差2种排名方式,确定当月价格异常农产品和确定去年价格异常,今年价格仍然异常的农产品的判定方法。目前与农产品价格有关的信息网站大多只提供了原始价格信息显示,针对规划好的从不同网站抓取出的价格数据利用企业级工作平台MyEclipse开发出了农产品市场价格Web信息分析系统,实现了价格查询、价格分析、价格预测和价格监测等功能,价格分析功能包括价格走势、各省对比、品种对比、同比环比和市场对比,价格预测功能包括单值预测和趋势预测,各省对比包括某一天对比和某一段时间对比等,分析结果以折线图、柱形图或地图的形式显示,界面美观,功能实用。可满足农业管理部门、农业企业、农户准确掌握不同地区、不同农产品价格的变化动态与走势的需要。
[Abstract]:In recent years, the abnormal fluctuation of the market price of agricultural products in China has become the focus of social concern. The price fluctuation of agricultural products will not only have a direct impact on the farmers' income and the enthusiasm of the farmers, but also to the daily life and the vital interests of the people. In order to maintain the stable and healthy development of the economy, to guarantee the living of the masses, the macro-control of stable prices is particularly important. Price monitoring and forecasting are an important link in the maintenance of price stability. The market price of the agricultural products in China is wide, the update is fast, it is urgent to establish a set of vertical search engine system to realize the data of the price of the agricultural products in the network on a regular basis, and an analysis, prediction and monitoring platform for the price of agricultural products is urgently needed to provide a comprehensive and clear analysis result. To provide the basis of production and control and decision-making analysis for the management of government departments, to become the decision-making basis for farmers to plant the plant, and to make a positive contribution to the market price stability of the agricultural products. The price information of the agricultural products provided by the price information network such as the agricultural net is not uniform, the product name is not standardized, and the like, and the analysis and the summary of the price data of the agricultural products distributed on different websites are analyzed and summarized, and the product name, the product market and the like are put forward. Name, initialize the agricultural product category, initialize the provinces and cities, standardize the data units, go to the re-duplication policy selection and the zero-price data processing for 7 data standardization The principle is to study the DOM tree method, regular expression, HTMLParser to extract the text information of the web page, and set up the price vertical search engine system of the agricultural product market by using the software of Heritrix and so on. The price information of different non-agricultural products can be extracted by using the software of Heritrix and the like, and the price of the agricultural products of the whole SQL server is formed after the standardization. according to the invention, in order to further improve the forecasting precision of the market price of the agricultural products, the agricultural products with abnormal prices are found in time, and the carrot, the white radish, the Chinese cabbage, the scallion, the bean horn, the cucumber, the pepper, the leek and the eggplant are selected, Prediction of thirteen agricultural products of green pepper, potato, tomato and rape In the weighted arithmetic average prediction method, the method of five weights is compared and analyzed, and the experimental results show that the method is superior to the other weight setting method in the current year price, which is better than the simple method in the average prediction method. The results show that the trend continuation method of the quadratic curve and the trend continuation method of the time series are not applicable to the prediction of the price of the agricultural products. In the simple arithmetic mean method, the weighted arithmetic mean method and the time series average increase the long-volume prediction method, the time series geometric mean method, the one-time moving average prediction method, the secondary moving average prediction method, the primary index smoothing method, the quadratic exponential smoothing method and the straight-line trend continuation method The method is applicable to the prediction of the price of agricultural products, and on the basis of that, an improved quadratic exponential smoothing prediction method, a secondary exponential smoothing prediction method and a secondary exponential smoothing prediction method are provided, The error sum of the errors is the lowest, the improved quadratic exponential smoothing method is better than the last modified quadratic exponential smoothing method, which is superior to the quadratic moving average prediction method, and the time series seasonal index method and the seasonal index trend method are compared, and the experimental results are obtained. It can be seen from among the two prediction methods of the most agricultural products that the sum of the errors of the errors is much different, the actual predicted price can be the predicted value with low error square, and the historical error 2 from the predicted value and the actual price error is proposed in the aspect of the determination of the abnormal price of the price. In a way of ranking, it is determined that the price of the current month is abnormal and the prices are abnormal last year, and the prices at this year are still abnormal. At present, most of the information websites related to the price of the agricultural products only provide the original price information display, and the price data analysis system of the agricultural product market price is developed by using the enterprise-class work platform MyEclipse for the planned price data from different websites. The function of price query, price analysis, price forecast and price monitoring is realized. The price analysis function includes the price trend, the comparison of each province, the comparison of the varieties, the year-on-year link and the market comparison, the price prediction function includes single-value prediction and trend prediction, and the comparison between the provinces includes one day's comparison and the trend prediction. For a period of time, the results of the analysis are displayed in the form of a line chart, a bar graph, or a map, and the boundary The surface is beautiful, the function is practical, the agricultural management department, the agricultural enterprise and the farmer can accurately grasp the change of the price of different agricultural products
【学位授予单位】：沈阳农业大学
【学位级别】：博士
【学位授予年份】：2013
【分类号】：TP391.3

【参考文献】