1.  Introduction

In the loan behavior of individual consumers and small and micro enterprises from financial institutions, the problem of bad debts urgently needs to be solved. Information asymmetry makes it impossible for financial institutions to promptly understand the true financial situation of borrowers and determine reliable credit ratings. Moreover, the low entry threshold of the white list makes it difficult for borrowers whose actual credit does not meet the standards to repay on time after successfully borrowing. This led to a further increase in the bad debt rate [1].

Today, with the increasing maturity of big data technology, financial institutions have valuable risk identification tools. However, financial institutions themselves are still in the early stage of technological development and encounter difficulties in systematically and comprehensively identifying and managing risks [2]. The organization, integration and analysis of massive correlated data by big data technology can greatly assist financial institutions in efficiently rating borrowers' credit ratings, serving subsequent decision-making and possible risk hedging strategies [3]. However, due to the problems of big data such as privacy leakage and insufficient quality inspection [4]. In the collection, processing and use of data. The protection of user privacy, as well as the sources of data, data screening during the data processing and supervision by relevant departments, all need to be carefully considered.

This study aims to explore the role that the high efficiency, high precision and high relevance of big data technology can play in the credit assessment of borrowers and the risk identification brought about by financial institutions based on this assessment. However, given the current emphasis placed by the Internet finance industry on rural revitalization and inclusive finance, highlighting "prosperity" and sharing, financial development also shoulders the mission of providing innovative vitality to underdeveloped regions and enhancing people's well-being [5]. When dealing with borrowers from rural areas, the rating can be appropriately downgraded to protect the rights and interests of financial institutions while maintaining humanistic care.

2.  Theoretical basis

Big data technology refers to the collection, cleaning, analysis, storage, etc. of data from different sources, including structured data and unstructured data. It features high speed and high accuracy [6]. Among them, the "analysis" stage is the key to making efficient and correct decisions. In the field of credit reporting, big data is often used to collect borrowers' credit reports, housing provident fund contribution data, etc. Because it can discover other data of borrowers, such as social media and customer transactions, and establish correlations with their credit reports, it has also become an important application to formulate personalized loan plans, segment borrower groups and implement different communication strategies [7]. For small and micro enterprises, debt restructuring and bound operation improvement should be adopted, and an exit mechanism should be designed. For individual consumers, measures such as stress-relieving communication and income reconstruction plans should be adopted. The conclusion of the communication strategy is based on data analysis. The following table 1 shows several data analysis frameworks.

Data analysis is a process that studies structured and unstructured data and generates useful analyses through statistical and quantitative analysis methods, with the aid of inferential models [8]. Through data analysis, financial institutions can assess the repayment risks of borrowers in advance (predictive analysis), infer possible causes (diagnostic analysis), provide risk hedging strategies (normative analysis), generate relevant reports, etc. (descriptive analysis).

Data analysis can monitor in real time and integrate multi-dimensional data to detect financial fraud (such as identity theft, account takeover, etc.) [9]. It can not only help financial institutions identify the risks of borrowers maliciously defaulting on debts and failing to repay on time. For borrowers, it can also help them protect their accounts, financial and identity information from being misused, effectively safeguarding their privacy [10].

3.  Factors of loan risk

Credit risk is classified into the following types: default risk (borrowers fail to repay on time), concentration risk (excessive exposure to a single entity or borrower), systemic risk (chain reactions triggered by institutional turmoil or default by financial institutions), and other types of risks. This study mainly focuses on default risk, takes concentration risk into account, and does not involve systemic risk or other types of risk, as can be seen in table 2. After research, the influencing factors affecting borrowers' repayment are as follows [11, 12] .

Summarizing the factors of debt risk, this article conducts a more detailed exploration of social acceptance. The following table 3 shows the acceptance of big data credit investigation by different regions and age groups.

Past studies have shown that big data can reduce credit analysis errors by 40%. However, the proportion in Uzbekistan is still around 15% to 20%. For developed countries, the accuracy rate of credit analysis will also lead to an increase in the proportion of trust in numerical scores [8].

4.  Efficiency of loan risk assessment

A series of past studies have analyzed the project materials and customer surveys of the Central Bank of Uzbekistan. By using credit risk ratios, artificial intelligence models and SWOT analysis methods, etc., the efficiency (including time, cost, error rate, etc.) of traditional assessment and digital assessment is compared.

(Data sources: Report of the Central Bank of Uzbekistan (2024), data from the Industrial Construction Bank of Uzbekistan)

According to the table 4, it can be seen that the efficiency, fault tolerance rate and cost of digital loans are all significantly lower than those of traditional methods. Moreover, customers are more satisfied with digital loans and call for further development of digital loans. It can be seen that relying on big data technology, the loan services of financial institutions have been greatly improved, which is undoubtedly a positive factor for identifying and further reducing risks.

Research has found that if a central data platform is established, the accuracy rate of scoring will increase by 30% to 35%. Legislative reform can increase customer trust by 20% to 25%. This result strongly proves that a central data platform should be established, with classified and organized data, and relevant data should be connected to obtain more convincing credit scores. At the same time, legislation should be enacted to clarify how social media data is used and to what extent. While fully protecting the borrower's privacy, a more personalized borrowing plan is also specified [8].

Uzbekistan is a less developed country. The overall financial literacy of borrowers is relatively low, and their debt-paying ability is also low. Considering the limitations of less developed countries like Uzbekistan and rural areas, the principle of inclusive finance should also be kept in mind. The dissemination of digital inclusive finance. By investigating the production investment, household income and the strength of social relations (quantifying the number of gifts given in the past) of tens of thousands of rural Chinese families, farmers with strong social relations were identified. Through mutual information exchange, they learned about the benefits of digital finance and enhanced its influence. With the increase in the time spent on mobile financial services, farmers' financial knowledge has improved, and the utilization of digital finance has increased productive investment, leading to an increase in the number of loans [13]. Therefore, through digital inclusive finance, borrowers in rural areas can enhance their financial literacy and contribute to financial development [14].

5.  Conclusion

Big data can be classified into four types: descriptive analysis, diagnostic analysis, predictive analysis, and normative analysis. In the field of financial credit information, descriptive analysis is often used to generate credit reports for customers, diagnostic analysis is employed to explore the reasons why borrowers failed to repay on time in the past, and predictive analysis is utilized to predict the financial situation of each customer and pre-arrange the possibility of repayment, and to formulate hedging strategies. Determine the action plan for communicating with the borrower and whether to grant a loan through normative analysis.

To ensure that big data technology precisely empowers the credit reporting field and effectively protects the legitimate rights and interests of financial institutions, we need to pay attention to the following factors. The factors include historical model factors (such as credit scores, income levels, and employment history, etc.), legal and infrastructure issues, social acceptance, and the borrower's own personality. In many of these developing countries, the laws regarding the use of social media data are still unclear, and the infrastructure is still not well-developed. There is a significant difference in social acceptance among people of different ages and income levels (only data from Uzbekistan is provided, with a comparison of the world average data). Whether a borrower's own personality is overly optimistic or not will affect their judgment of the possibility of timely repayment, and thus influence their credit score.

In regions with different levels of development, there are differences in the efficiency and satisfaction of big data Windows. In developing countries represented by Uzbekistan, the credit period limit, error rate, cost (per loan), and customer satisfaction of the digital window are all far superior to those of traditional methods. There is an urgent call for the promotion of digital financial credit reporting. And considering that the accuracy rate of big data analysis in Uzbekistan is lower than the world average. In developed countries, as the accuracy rate improves, the efficiency and satisfaction of digital Windows will only increase.

To better enable big data technology to serve the credit investigation and financial fields, the following suggestions are hereby put forward. Necessary financial knowledge should be taught to borrowers to ensure they have a correct understanding of sudden income fluctuations and the probability of bankruptcy and report them in a timely manner. Establish a central data platform, carry out legislative reforms on social media, clearly define the boundaries for financial institutions to use borrowers' social media data, and provide personalized borrowing plans while protecting borrowers' privacy and security. Enhance the accuracy and efficiency of big data analysis technology, teach financial knowledge through digital loans, and increase trust in digital finance and credit scoring. Guide small loans, taking into account the characteristics of unstable income and lack of financial knowledge, and flexibly adjust the repayment period for rural areas.

This study only collected partial data and did not conduct a unified statistics on the trust level of people around the world in digital finance. There is a risk of overgeneralization. Therefore, the world average level was compared to enhance credibility and persuasiveness. At the same time, it mainly focuses on the perspective of financial institutions and does not compare the differences among various digital technologies, resulting in omissions of technical issues. In the future, we can focus on the advantages and disadvantages of different digital technologies, and combine the exploration of systemic risks and other risks to provide a new paradigm for breakthroughs in the digital lending field.