Based on CiteSpace Knowledge GraphAnalysis of the Status Quo of Research on Expenditure Forecast

This paper uses CiteSpace software to conduct text mining and visual analysis of 744 academic papers related to funding forecasts in the CNKI database in the past 20 years, sort out the literature time, journals, authors, and institutions of funding forecasts, and use keywords co-occurrence and subject terms Clustering and other methods reveal the current research status and evolutionary laws of funding forecasting, hoping to provide reference and reference for the selection of other types of funding forecasting methods.


Introduction
This article uses the CiteSpace information visualization tool, takes the journals included in the CNKI database as the object, and takes the period from 1980 to 2021 as the time range, and conducts an indepth analysis and analysis of the funding forecast related literature from the aspects of keyword cooccurrence, cluster analysis and hot word emergence Comment, grasp the status quo and laws of development, summarize research deficiencies, predict development trends, and provide reference and reference for the selection of other types of funding forecasting methods.

Analysis tools
This paper uses the information visualization software CiteSpace developed by Professor Chen Chaomei from the School of Computer and Information Science of Drexel University to conduct text mining and visualization analysis on the research topic of "expenditure prediction" [1]. CiteSpace is the abbreviation of citation space. CiteSpace is a multivariate, time-sharing and dynamic citation visualization analysis software that focuses on analyzing the potential knowledge contained in scientific literature. It can extract useful information from a large amount of literature and convert it into visual scientific knowledge. The map can visually present the laws hidden in a large amount of data, and effectively display the network, structure, interaction, evolution or derivation of many complex relationships between knowledge units [2][3].

Data sources
Search the CNKI database according to the subject containing "expenditure forecast" or "expense forecast". As of February 24, 2021, a total of 887 papers have been retrieved. Non-academic papers such as conferences and press releases or articles irrelevant to the topic are excluded. Secondly, the selected data is exported in Refworks format, named as a CiteSpace recognizable name, and filtered, deduplicated and converted by the data converter, and finally a total of 744 valid data are obtained.

Analysis of Annual Paper Volume
The number of documents is an important indicator to measure the development of a certain subject or topic, and it is also one of the important scales of academic research attention. According to Figure 1, it can be seen that the research of funding forecast is directly affected by the needs of actual tasks and the degree of development of machine learning and various algorithms.

Distribution of articles in journals
Academic journals are an important link and an important carrier of academic communication, as well as the test of academic research results. By analyzing the source journals of the literature, it is helpful to determine the main source of the research field, and to a certain extent reflect the research in this field. Degree and so on [4].

Co-occurrence analysis of research authors
This article uses the micro-scholar cooperation network provided by CiteSpace to find out the cooperation between authors in the field. In the co-occurrence network of authors drawn by CiteSpace, the size of the node represents the amount of literature published by scholars. As shown in Table 2, the author's co-occurrence network contains a total of 683 nodes (N) and 582 edges (E), and the density (Density) is 0.0025. It can be seen that although a total of 683 scholars participated in the completion of 582 academic papers, the connections between scholars are not close, and they are still mainly individuals or small groups. According to "Price's Law", the expression formula for the number of articles published by the core author is max 749 . 0 Np Mp  (1) Among them, Np max is the number of articles published by the most productive scholars in this field [5]. According to the results of the CiteSpace software, Zhang Hengxi has published 13 papers, ranking first, so the minimum number of published papers for core authors is 3. From this, There were 50 core authors from 1980 to 2021. It can be seen from the figure that the distribution of funding forecast research institutions is affected by the nature of disciplines and units. It can be seen that In the field of engineering disciplines, the main problem solved by funding forecasts may be of a military and national defense nature. However, according to the knowledge graph, it can be seen that organizations with strong connections are often in the same department, and there is still less cross-departmental cooperation.

Fig 3. Keyword cluster co-occurrence analysis
Based on the results of the cluster analysis, the research topics of funding forecasting are finally summarized into the following two aspects: First, research on the application of funding forecasts. Key words include medical expenses, total health expenses, research and development expenses, funding requirements, medical expenses, treatment expenses, etc. The second is the study of specific methods of funding forecast. Key words include GM (1,1) model, regression analysis, combination forecasting, grey system, quantitative analysis, etc.

Keyword emergence analysis
Use the mutation test function of citespace to get the keyword emergence map in this field, so as to understand the research frontier of this field. Based on the cluster analysis, this paper uses citespace to sort the keywords according to the length of the keyword emergence duration, the emergence start time, and the emergence intensity.
Based on the analysis of the emergence time length, emergence intensity, and emergence start time, the most prominent methods are support vector machines and ARIMA model methods. When making expenditure forecasts, a combination of forecasting methods can be used to combine the two to obtain The best prediction result.

Analysis of Research and Development Process
Keyword clustering sequence diagram is used to describe the historical evolution process of a certain cluster and the relationship between each cluster. Combining clusters with similar meanings together, cluster #0 (expense prediction), cluster #1 (prediction), and cluster #3 (prediction model) have continued to be research hotspots since 1986 , The attention so far has not decreased. Cluster #2 (total health expenses), cluster #4 (hospitalization expenses), and cluster #5 (medical insurance) have been paid attention to by experts and scholars from 1987 to 2021, and they are still applying funding forecasts The field. The research on cluster #6 (model), cluster #7 (gray prediction), and cluster #9 (gray system) started in 1990, and the research continued until 2018.

summary
Research on funding forecasts has been an academic hotspot since 1986. Its application fields and covered subjects were initially mainly in the medical field and engineering equipment fields, but now they are more and more widely used in economics, education, sports, transportation, computers, etc. each field. With regard to the choice of the method of expenditure forecasting, according to the comprehensive analysis of the visualized results of citespace, the most promising and more mature methods are the support vector machine and ARIMA model.
Although funding forecasting research has always paid more attention and the scope and fields of application are more extensive, there is a lack of innovation in the selection of specific methods. Most of the widely used forecasting methods have been applied for more than 20 years and have universal applicability, but they are different. Funds have different characteristics. Therefore, when making fund forecasts, it is necessary to improve on the basis of existing forecasting methods in order to obtain more scientific and accurate forecast results.