Xinling GAN, Yong LI

Institute of Information Engineering, Binzhou University, Binzhou 256600, China

Abstract This paper explains the basic concepts of "Internet+" and big data, analyzes the main problems in the application of big data technology in agricultural informationization of Shandong Province, summarizes corresponding solutions from the aspects of government guidance, financial input, open sharing of agricultural big data, big data storage and processing, data mining, etc., and prospects the application trend of big data technology in agricultural informationization to achieve the connotative development of agriculture in Shandong Province.

Key words Agricultural informationization, "Internet+", Big data

1 "Internet+" and big data

1.1 Concept of "Internet+"In July 2015, theGuidingOpinionsoftheStateCouncilonActivelyAdvancing"Internetplus"Actionwas issued. As a further practice of Internet thinking, the "Internet+" action boosts the evolution of the economic shape and drives the vitality of social and economic entities, thus providing a broad network platform for reform, innovation and development.

As an important driving force for agricultural transfer and restructuring, "Internet+agriculture" boosts modern agriculture to a technology-intensive, data-intensive transition through introducing Internet-based infrastructure, smart equipment, and data services in agricultural production management and management services, thus improving comprehensive agricultural production capacity. The "Internet+agriculture" applications mainly include "Internet+agricultural Internet of things" to achieve agricultural intelligence, "Internet+agricultural e-commerce" to achieve agricultural management network, "Internet+agricultural information services" to achieve information into village, and "Internet+agricultural big data" to achieve forecasting and early warning[1-3].

1.2 Concept of big dataBig data refers to the data of which the time used for collection, processing and analysis by commonly used software is much longer than that humans can tolerate. The industry usually classifies the characteristics of big data as "3V", also known as 3V attributes, namely, volume (large volume), variety (multiple varieties) and velocity (high velocity). Agricultural big data refers to big data technologies that are applied, practiced, and developed in agriculture and related fields. The agricultural sector contains the soil that big data produces. At the same time, it can also provide a broader stage for the application of big data technology. This includes agriculture, forestry, animal husbandry, sidelines, fishing and other agriculture-related industries, as well as cross-professional and cross-industry, even cross-industry data collection, data analysis, data integration and data mining. Decision support and decision management are also included. Big data in agriculture field has its inherent characteristics and can be summarized in two aspects,i.e., complexity and imbalance. First, the data is complicated. Second, the data is imbalanced[4-6].

2 Main problems in the application of big data in agricultural informationization

In June 2013, Agricultural Big Data Industry Technology Innovation Strategic Alliance was established in Shandong Agricultural University. It is the first institution in China to research and apply agricultural big data, achieving a good docking of agricultural big data research and application. However, agricultural big data faces the technical and policy barriers such as wide heterogeneity, incompleteness, real-time processing, lack of prior knowledge and privacy. Therefore, agricultural big data still needs to go forward. In the following five aspects of agricultural informationization infrastructure investment, the main problems in the application of big data in agricultural informationization are analyzed and summarized.

2.1 Infrastructure construction of agricultural informationization needs to be further strengthenedBecause of the gap between economic opening and development, the degree of informationization construction between different urban and rural areas and different regions is obviously different. At the same time, different agricultural informationization development concepts and economic levels have led to the emergence of the "Matthew effect". It is more difficult to coordinate and develop agricultural informationization in a coordinated manner. It is necessary to further increase the input into the agricultural information infrastructure such as the power grid, telephone and Internet. According to incomplete statistics, in 2012, the computer holdings of urban residents were 87 sets/100 households, and the computer holdings of rural residents were 21.4 sets/100 households. According to China’s 33rd Internet development statistics, rural netizens are gradually increasing in China. As of 2013, the number of rural netizens increased to 177 million, accounting for 28.6% of all Internet users in China. However, the Internet penetration rate in rural areas is only 27.5%. Computer education has a lower penetration rate in rural areas. Currently, the popularity of mobile smart terminal devices has vigorously promoted the development and popularization of networks in rural areas. However, the application of mobile networks in rural areas mostly focuses on leisure and entertainment. Education support for rural big data development and utilization and universal talent training has not yet been provided.

2.2 Lack of professional and technical personnel and policy supportThe establishment of an interdisciplinary composite team is an important guarantee for the research and application of big data. The formation of agricultural big data talent team should start from agricultural technology experts to agricultural product sales consultants, and from agricultural information chemists, statistical analysis talents to network engineers. At present, in Shandong Province, the construction of interdisciplinary composite big data talent team is backward, the talents in agricultural big data research, application and management are insufficient, there are not enough ways to train talents, and the identification and evaluation criteria for full-time employees of agricultural big data have not yet been constructed. The formulation of laws, regulations and institutional standards for openness and sharing of data is relatively backward. Detailed specifications such as sharing and openness principles, data formats, quality standards, usability and interoperability have not yet been formed. The data sharing and openness of government departments and public institutions is not strong, the level is not high, and quality is poor.

2.3 Lack of unified management standards and low level of information sharingData sharing issues have not received enough attention due to historical reasons, leading to differences in format, type, and storage standards of historical data. First, data sharing techniques in different regions, different industries, and different fields have not yet formed sharing standards. Agriculture multi-source heterogeneous, structured and semi-structured data standardization and massive data management model technologies are particularly lacking. Second, there is a lack of cross-platform integration technology. Various websites and data platforms have been built in the process of agricultural informationization. The level between the platforms is not clear enough, and the coverage content is inconsistent. As a result, interconnection and intercommunication between websites and platforms cannot be achieved, and information silos are serious. Third, the convenient and efficient data query, browsing, retrieval and distribution technologies have not yet been established.

2.4 Data analysis methods and processing models need to be improvedBecause agricultural big data is mainly generated from IoT RF equipment, agricultural information websites and various mobile terminals, the collected data is transferred from typical structured data to structured, semi-structured and unstructured fusion data. How to achieve unified storage and analysis of converged data needs to be solved. The knowledge value of data is strong in real time and decays with time. Therefore, data timeliness is particularly important in the process of agricultural big data analysis. For example, the timeliness of weather and environmental conditions and other related data analysis directly determines the effects of agricultural production disaster management. The analysis of the specified amount of data in a tolerable time is an important indicator for measuring big data analysis, and the data processing mode combining stream processing and batch processing is adopted for dense data volume.

2.5 The ability to mine agricultural big data needs to be further improvedHeterogeneous storage of agricultural data leads to diversification of data types. The large and complex data sets make traditional machine learning and data mining algorithms no longer suitable for mining agricultural big data. The existing data mining and machine algorithms are mainly for small-volume data. If they are applied to large data sets, the analysis results may have large error or even unusable. Cloud computing will be an important tool for agricultural big data processing. In addition, the real-time nature of agricultural data applications is particularly important, and the accuracy of data processing algorithms may be second. It needs to find a balance between the real-time and accuracy of data mining and processing[7-10].

3 Application of big data in agricultural informationization

Shandong Province should seize the opportunity of agricultural information development, actively play the guiding position of the government, and introduce specific policy measures and implement them to promote the use of agricultural big data. It should also increase government financial investment, gradually establish the openness and sharing of agricultural big data, promote resource integration, break down data barriers, update big data storage methods and processing modes, and improve data mining algorithms to promote the application of big data in agricultural informationization.

3.1 Playing the government’s guiding positionIn order to better coordinate the work of agricultural big data in Shandong Province, the Shandong Agricultural Big Data Work Leading Group was established to uniformly lead the application of agricultural big data in Shandong Province. The team leader fully coordinated the Shandong Provincial Development and Reform Commission, the Economic and Information Technology Commission, Department of Science and Technology, Department of Finance, Department of Agriculture, Department of Ocean and Fisheries, Department of Forestry, Animal Husbandry and Veterinary Bureau, Agricultural Machinery Administration, Academy of Agricultural Sciences, Shandong Agricultural University and other relevant departments. The office is located in the Agriculture Department of Shandong Province, responsible for implementing specific measures and daily work to promote the use of agricultural big data. A pattern of "one department, one voice" is formed. Sharing is promoted through openness. The municipal, county (city, district) governments and agricultural administrative departments shall refer to the practices in the province, and introduce specific policy measures and implement them in conjunction with local conditions, thus promoting the use of agricultural big data.

3.2 Increasing government financial inputThrough the establishment of special funds for agricultural big data, provincial finance guides local governments, industrial and commercial enterprises and competent and demanding agricultural production and management entities to actively participate in. Through implementing financial subsidies, tax reductions and other preferential policies for social investment in the field of agricultural big data, a government-led, enterprise-invested, full-participation diversified investment pattern has been gradually formed. It is recommended that provincial finance set up special fund support in the construction of agricultural big data exchange centers, the development of data exchange standards and the development of big data application software. Agricultural big data application demonstration counties and key counties are established. A special plan for the cultivation of agricultural big data talents is formulated to train high-level talents in agricultural big data and gradually establish a high-level talent team that adapts to the promotion of agricultural big data.

3.3 Increasing the openness and sharing of agricultural big dataBig data is used for data openness. The value behind complex data is exploited to turn it into a business model. New data such as weather forecasts, soil conditions, GPS maps, water resources, market environment, market demand,etc., counted by government departments, have not be disclosed or opened, limiting user access. Therefore, the government and relevant departments should actively promote agricultural data information for data opening. In addition, China should speed up the development of a widely recognized and adopted data openness national standard, thereby promoting the development and utilization of agricultural data.

3.4 Updating the big data storage mode and processing modeStructured, semi-structured and unstructured data are the primary forms of agricultural big data storage. To achieve subsequent storage and analysis of agricultural big data, it is necessary to simplify the conversion of complex multi-type agricultural big data. The converted data may have errors, valueless or even distracting data. Therefore, the data needs to be filtered. For massive data that focuses on real-time, fast-call requirements, relying on traditional data storage methods, cloud storage technology and data development integrated platform under Hadoop and Mapreduce platforms can be developed to run data storage and computing functions concurrently. For managing multi-source, cumbersome and massive data files, new computing technologies should be developed. Fast call of massive image data and dynamic display of fuzzy research and accurate search operations are supported.

3.5 Improving data mining algorithmsThe traditional data mining algorithm of machine learning has not adapted to agricultural big data analysis. This is because that algorithm suitable for mining small-volume data cannot be applied to big data statistics and analysis. In addition, due to the characteristics of agricultural big data, the accuracy of mining algorithms is no longer the main indicator. In most cases, the timeliness and processing accuracy of the mining algorithm need to be balanced. The goal of big data analysis is not to analyze the data, but to get valuable information by statistical analysis. For example, after forming and improving the data of the whole industry chain of pigs, through comparison and in-depth mining of data at various stages of the industry chain, the core issues of the current hog industry development are gotten, the regularity information of market supply and demand changes is obtained, thereby facilitating the formulation of a more scientific "data-driven" pig regulation policy[11].