Intelligent financial decision support system based on big data

3.1 Web big data crawler technology and ETL data process

The user gives instructions through the computer, and the web crawler technology starts acting according to the accepted instructions and automatically collects the target data [11]. Web crawler technology has the characteristics of real-time collection and automatic collection, the former guaranteeing the timeliness of the data and the latter guaranteeing the accuracy and integrity of the data, which is conducive to the diverse storage and analysis of the data [12]. In order to improve resource utilization, this study proposes a web scraping algorithm and applies it to a financial decision support system. During the use of web crawling technology, important information on web pages should be extracted. The study uses the best-first search method to judge topic similarity.

where term frequency (TF) refers to the frequency of a specific word in the selected text, which is mainly used to increase the weight of keywords. n i , j indicates the occurrence frequency of a word t i in the text d j , so the sum of all words in the text can be represented by ∑ k n k , j .

where IDF denotes the reverse document frequency, which is used to reduce the topic weight of public words [13]. ∣ D ∣ denotes the total number of corpus documents, and the number of documents containing the word t i is denoted by ∣ { j : t i ∈ d i } ∣ .

where TF × IDF refers to the weight w i , j of the word, which is mainly used for the ranking of weights, while obtaining Top-N, which acts as the keyword of the text. The calculation of topic relevance is performed by the vector space model algorithm, so that the vector space dimension is the number of keywords n. At this time, the weights of different keywords k are the values of the corresponding different dimensions [14].

where the weight of specific keyword k in topic q is represented by f k , q . If p indicates the page, the keyword with the highest frequency of occurrence is selected as the target, and the corresponding frequency is expressed by x i = 1 .

which is the topic-defining equation of page p , where x 1 w i is the value of a single dimension in the corresponding vector within the page.

Formula (6) is an expression for the usage harvest rate, which is the ratio of the number of topic-related pages to the number of pages that have been extracted. This index is mainly used to judge the relevance of pages. The meaning of each letter in it represents the same as the description of the above formulas. Considering the real situation, a threshold r is set, and when cos 〈 α , β 〉 ≥ r , the page is relevant to the topic.

In the ETL of data warehouse, “E” refers to “Extracting data,” “T” refers to “Transforming data,” and “L” refers to “Loading data.” The ETL also includes “Cleaning “ which means “Cleaning data” [15]. Data warehouse can be divided into client layer (querying and customizing reports), server layer (filtering system junk data, transforming data formats), and online analysis server layer (simplifying processes and increasing data transfer quality) [16].

Figure 1 shows the schematic diagram of the ETL architecture of the data warehouse of J Group company. The source data system includes enterprise accounting system, reporting system, financial sharing system, UFIDA software, Kingdee software, and Longchao software. After the data are extracted, transformed, cleaned, loaded, and other series of operations are completed, they enter the data warehouse sharing platform of the group’s operation distance vehicle together with the external data obtained through the network big data crawler system. If the data are stored in a relational database and the data are structured, the extraction of the data is completed using SOL statements, and the target data are extracted into a temporary database through metadata mapping rules, combined with a SQL language generator, and processed and output to the target database [17,18].

Figure 1

ETL architecture diagram.

3.2 Intelligent financial decision support system architecture

Intelligent financial decision support system is a technology that combines computer artificial intelligence with management science. It achieves the construction of a working environment with a combination of knowledge, initiative, and information processing capabilities by means of artificial intelligence dialogues to provide decision support work for corporate decision makers and substantially improve the management of dry-level managers [19,20]. The study selected 50 enterprises in province S for analysis, 36 of which belong to groups, 7 to shares/units, 3 to group finance companies, 3 to administrative institutions, and 1 is of banking nature.

Table 1 shows the demand analysis results of 50 enterprises in S province for intelligent financial decision support system. The above 50 enterprises have the strongest demand for data analysis, followed by demand for industry benchmarking, demand for processing and analysis of external big data, demand for smart financial analysis report, demand for risk warning, demand for artificial intelligence modeling, demand for intelligent interaction, and demand for self-service. Among them, industry benchmarking is mainly for the comparison of financial data in the same industry, and intelligent prediction is mainly for the prediction of enterprise profit and enterprise capital structure.

Figure 2 shows that the intelligent financial decision support system contains components such as multi-source heterogeneous data layer, application scenarios, and operating decision data sharing platform. The application scenarios are built by the expert reasoning and financial intelligence support platform, where the modeling tool platform includes the reporting engine, interaction engine, dashboard engine, and indicator engine. Based on the data processing technology to generate financial analysis portal, the human-computer interaction layer is established so that the top management of the enterprise can understand the operation status of their own enterprise in a timely, comprehensive, and accurate manner and help the top management to make decisions.

Figure 2

Intelligent financial decision support system architecture.

4.1 The operational effect of intelligent financial decision support system

The experiment takes J Group in S province as an example. After investigation, the problems in J Group’s financial management include: first, too many forms, which are not easy to manage. Second, the data are easy to be lost and leaked. Third, the lack of in-depth analysis of financial data. Fourth, some information is entered several times, lacking reference to the same industry data. Fifth, cannot automatically generate charts and text as well as the corresponding enterprise financial management reports. In the architecture of intelligent financial decision support system, the effective mining of external data is realized through big data web crawling technology, and the crawling process is presented through visualization tools, see Figure 3 for details.

Figure 3

Real-time monitoring of external data.

As shown in Figure 3, with the date as the horizontal coordinate and the data volume as the vertical coordinate, it can be seen that the data collected by the big data web crawler technology are in a state of real-time change. The crawler technology collects and shows the changes in the enterprise finance-related data, including basic data, industry big data, Internet of Things (IOT) data, OA system data, and many other external data, which can supplement the contents of the corresponding enterprise’s decision support items and improve the accuracy of decision support.

As shown in Figure 4, multi-dimensional analysis of financial data is completed by OLAP technology. The bar chart is used to display the core indicators of the enterprise such as profits, return on net assets, and accounts receivable. The dashboard is used to display the budget situation of different indicators of the enterprise. Using drill-down to establish the correlation between data and drill-down by company hierarchy, it is convenient for company management to query the status of corporate financial data and conduct linkage queries on the causes of financial variances hidden behind the financial data. Managers can click on the bar chart to view the specific situation of different segments of the enterprise such as operating income, return on net assets, and budget execution rate.

Figure 4

Partial operation results of J Group’s intelligent financial decision support system.

As shown in Figure 5, according to the intelligent financial decision support system, J Group’s gearing ratio reaches 55.27%, current assets growth rate reaches 10.38%, and operating income growth rate reaches 20.28%. Its financial expenses reach $19.74 million. Its return on total assets is at an excellent level in the industry rating, while the growth rate of total assets is 4.77%, which is at a lower level in the same industry. Overall, its current assets growth is slower than the growth of main business revenue, while the profitability of assets improved and the enterprise’s asset structure improves regionally. Managers of this enterprise should make financial decisions with the main purpose of increasing the growth rate of total assets and maintaining the total assets return of the enterprise in a targeted manner.

Figure 5

Intelligent prediction results of Group J.

4.2 Discussion of the operational effect

According to the test results of the system operation, the system proposed by the research is far better than the traditional operation mode of the company in terms of data mining capability. The data mining ability of the system proposed by the research institute is mainly realized through crawler technology. According to the test results, the daily mining volume of the system for basic data, industry big data, IoT data, and OA system data reaches 28 G, 26 G, 20 G, and 24 G, which is significantly higher than the traditional data means used by the company. According to the test results, the research institute proposes that the system has user interaction interface and data visualization capabilities. Users can obtain the desired data statistics results through simple operations. These results are displayed in the form of visual charts, enabling users to easily obtain the useful information contained in the data. The data display interface of the system contains multiple core financial indicators. The specific data processing results page can be further expanded by users.

4.3 Application effectiveness of intelligent financial decision support system

By comparing the changes in work efficiency of the group before and after applying the intelligent financial decision support system, the practical application effect of the intelligent financial decision support system designed in this study is analyzed, as shown is Figure 6.

Figure 6

Changes in the efficiency of finance staff before and after the system is on. (a) Number of financial reports received by the Group per month. (b) Monthly time for the Group to complete analysis.

After adopting this intelligent financial decision support system, the number of financial reports received by the enterprise from its subsidiaries per month has increased significantly, and the length of time the enterprise takes to complete the analysis reports per month shows a significant decrease. It indicates that the financial analysis workload of the enterprise is reduced after the application of the intelligent financial decision support system. The application of intelligent financial decision support system also solves the problem of information asymmetry to a certain extent, which is conducive to the group to grasp the financial status of its subsidiaries in a timely manner and improve the control of the enterprise.

From Figure 7(a), it can be seen that this intelligent financial decision support system is applied to the benchmarking analysis of Group J. In terms of sales revenue growth rate analysis, the actual value of Group J is 0.63%, which is 31.27% less than the industry excellent value of 31.90%. The inventory turnover ratio, asset load ratio, and accounts receivable turnover ratio of Group J are lower than the industry excellent value. The interest multiplier earned and return on net assets of Group J are higher than the industry excellent value. Figure 7(b) shows the evaluation results of the profitability of the companies within Group J against the early warning through this system. It can be seen that the return on capital and return on net assets of Group J are at the excellent level, i.e., the return on capital and return on net assets of Group J have reached the level of the excellent value in the industry. Through this intelligent financial decision support system incorporating big data, enterprise managers can technically access relevant data, analyze moving indicators, address existing problems in a timely manner, take effective corrective actions, and ensure the sustainable and healthy development of the enterprise.

Figure 7

Evaluation of group financial indicators after the system is on. (a) Enterprise benchmarking analysis return on net assets. (b) Industry benchmarking radar chart of corporate profitability.

4.4 Discussion of the application effectiveness

There may be obvious deviation between the performance of a system in the running test and the actual application. Therefore, after the running test, the system needs to be placed in the actual application environment for testing. According to the changes in the work efficiency of the financial staff after the system went online, the number of financial statements received by the company each month increased significantly. In 2015, the Group received only 42 financial reports per month, and the analysis time of the report was up to 18 h. Since 2016, the number of financial reports received by the Group has exceeded 300, while the time of analysis has been declining. By 2019, the time of analysis has decreased to 2 h. According to the results, the system has significantly increased the amount of data analysis, while the analysis time has decreased, which shows that with the help of the system, the speed and efficiency of the analysts in processing information have increased. From the perspective of data analysis, the experiment then tested the evaluation of the group’s profitability after the system was launched. The results show that the system can display the financial analysis results of the group in a visual and numerical way, and the analysts can efficiently obtain relevant data and solve problems.

In summary, after the enterprise applies the intelligent financial decision support system designed by the experiment, the enterprise’s financial strategy can be better implemented, the enterprise’s management is optimized, the enterprise’s financial report generation time is shortened and can be generated intelligently, and the usefulness of the enterprise’s management decisions are also improved to a certain extent.