An intelligent recommendation method for financial technology products for small and micro enterprises

Abstract

The present invention discloses a method for intelligently recommending financial technology products for small and micro enterprises, which belongs to the field of financial service recommendation methods. It conducts in-depth analysis and graph construction in the financial field to make recommendations more accurate, break through traditional limitations, and adapt to the diverse and dynamic needs of enterprises. In cross-domain knowledge fusion, it integrates graphs through system acquisition and advanced technology to enhance application accuracy and provide multiple recommendation perspectives. It also optimizes product portfolios with quantitative evaluation models and optimization algorithms, considers key indicators and individual needs, dynamically adjusts parameters and strictly verifies them to ensure that the portfolio is effective and reliable in various market environments. Finally, intuitive recommendations, detailed explanations and feedback mechanisms enhance user experience, help the method to continuously improve, effectively promote the improvement of the quality and efficiency of financial services for small and micro enterprises, and promote their healthy development.

Description

Technical Field

[0001] The present invention relates to the field of financial service recommendation methods, and more specifically, to a method for intelligently recommending financial technology products for small and micro enterprises. Background Art

[0002] As small and micro enterprises play an increasingly important role in the national economy, meeting their financial needs is crucial to their survival and development.

[0003] In the field of fintech product recommendation, existing recommendation methods mainly recommend fintech products based on two aspects. On the one hand, starting from the financial field, they rely on the historical transaction data and financial statements of small and micro enterprises for analysis and recommendation. However, in the analysis and recommendation process, they have poor adaptability to fully consider the complex relationships between financial products and the diversified and dynamically changing financial needs of small and micro enterprises, and are relatively inconvenient to use. On the other hand, although cross-domain knowledge is applied to financial product recommendation methods, they lack systematicity and depth. For example, there are recommendations that simply analogize the concept of biological evolution, but they do not deeply explore the principles behind the complex relationship between species competition and cooperation in the ecosystem. As a result, the integration of knowledge in the financial field and the ecosystem field is not comprehensive and accurate when constructing a cross-domain knowledge graph, resulting in the knowledge graph being unable to accurately reflect the intrinsic connection between the two fields, which in turn affects the accuracy and effectiveness of the recommendation.

[0004] To sum up, the existing methods for recommending fintech products for small and micro enterprises have certain shortcomings and are unable to meet the growing complex financial needs of small and micro enterprises. There is an urgent need for a more scientific, comprehensive and accurate intelligent recommendation method to improve the quality and efficiency of financial services and promote the healthy development of small and micro enterprises. Therefore, we propose an intelligent recommendation method for fintech products for small and micro enterprises to solve the above problems. Summary of the Invention

[0005] In response to the problems existing in the prior art, the purpose of the present invention is to provide a method for intelligent recommendation of financial technology products for small and micro enterprises. It conducts in-depth analysis and graph construction in the financial field to make recommendations more accurate, break through traditional limitations, and adapt to the diverse and dynamic needs of enterprises. In cross-domain knowledge fusion, it integrates graphs through system acquisition and advanced technology to enhance application accuracy and provide multiple recommendation perspectives. It also optimizes product portfolios through quantitative evaluation models and optimization algorithms, considers key indicators and individual needs, dynamically adjusts parameters and strictly verifies them to ensure that the portfolio is effective and reliable in various market environments. Finally, intuitive recommendations, detailed explanations and feedback mechanisms enhance user experience, help the method to continuously improve, effectively promote the improvement of the quality and efficiency of financial services for small and micro enterprises, and promote their healthy development.

[0006] To solve the above problems, the present invention adopts the following technical solutions.

[0007] A method for intelligently recommending fintech products for small and micro enterprises includes the following interrelated and orderly steps.

[0008] S1. In-depth analysis and graph construction of financial field knowledge: For the financial field, systematically collect information on various financial products, including credit products, insurance products, payment and settlement products, and investment and wealth management products. For each type of financial product information, the interest rate pricing mechanism, repayment method, term structure, applicable small and micro enterprise industry types, enterprise size restrictions, risk rating system, and multi-dimensional characteristics of regulatory compliance requirements are sorted out in detail. Graph database technology is used to construct a financial field knowledge sub-graph. The nodes of the financial field knowledge sub-graph represent various financial products and their related attributes, and the edges of the financial field knowledge sub-graph represent the logical relationship between products and attributes, and between different products.

[0009] S2. Collection and structured representation of ecological domain knowledge: Systematically collect knowledge about species competition and cooperation in the ecosystem field, including the distribution of different species in ecological niches, resource acquisition strategies, energy flow and material circulation laws in ecosystems, and population dynamic change mechanisms. The resource acquisition strategies include predation, symbiosis, and parasitism. Use the web ontology language technology (OWL) to structure the knowledge of species competition and cooperation relationships and construct an ecological domain knowledge subgraph. The nodes of the ecological domain knowledge subgraph correspond to different species, ecological processes, and related ecological factors. The edges of the ecological domain knowledge subgraph represent the interactions between species and the associations between ecological processes and various factors.

[0010] S3. Fusion and association of cross-domain knowledge graphs: With the help of semantic analysis methods in natural language processing technology, including lexical semantic analysis, syntactic dependency analysis and semantic role labeling, we can deeply understand the semantic connotation of knowledge in the financial and ecological fields. On this basis, we use the knowledge association algorithm based on machine learning to establish a mapping relationship between the financial field knowledge sub-graph and the ecological field knowledge sub-graph, and integrate them to form a cross-domain knowledge graph. The cross-domain knowledge graph not only intuitively reflects the intrinsic connection between the knowledge of the two fields, but also provides a solid data foundation for subsequent knowledge migration and fusion.

[0011] S4. Analogy mapping of financial needs and ecological resources: The financial needs faced by small and micro enterprises are compared with the resource needs of species in the ecosystem. The various types of financial product information are regarded as different species in the ecosystem. Each type of financial product information meets the specific financial needs of small and micro enterprises through its unique multi-dimensional characteristics, that is, species obtain the required resources through their own survival strategies. The financial needs include financing needs, risk avoidance needs, and fund management needs, and the resource needs include food, habitat, and sunlight.

[0012] S5. Construction of a quantitative evaluation model for competition and complementary relationships: Based on the theory and mechanism of competition and cooperation among species in an ecosystem, a quantitative evaluation model for the competition and complementary relationships of financial product information is constructed. The quantitative evaluation model for the competition and complementary relationships of financial product information also fully considers multiple key attribute indicators of the financial product information to improve the accuracy of the construction of the quantitative evaluation model for the competition and complementary relationships. The key attribute indicators include profitability indicators, risk indicators, liquidity indicators, and cost indicators. The quantitative evaluation model for the competition and complementary relationships of financial product information also combines the personalized demand characteristics of small and micro enterprises, and through the multi-criteria decision analysis method of the analytic hierarchy process (AHP), assigns corresponding weights to the different demand dimensions of small and micro enterprises, and conducts a quantitative evaluation of the competition and complementary relationships of different financial product information in meeting the various financial needs of small and micro enterprises.

[0013] S6. Product portfolio search based on optimization algorithm: Using intelligent optimization algorithm to maximize the satisfaction of comprehensive financial needs of small and micro enterprises, minimize financial costs, diversify risks and stabilize returns as multi-objective functions, search in the cross-domain knowledge graph to generate potential financial product information combinations based on the cross-domain knowledge graph. The intelligent optimization algorithm includes genetic algorithm (GA), particle swarm optimization algorithm (PSO) and simulated annealing algorithm (SA), and during the search process, the intelligent optimization algorithm dynamically adjusts the composition parameters of the financial product information combination. The composition parameters include product types, product quantities and weight distribution of each product in the combination to adapt to the dynamically changing financial needs of different small and micro enterprises at different development stages and market environments.

[0014] S7. Rationality verification of product portfolio: A comprehensive rationality verification is conducted on the generated financial product information portfolio. The rationality verification includes carefully examining whether there are any conflicts or incompatibilities in the application conditions, usage period, operating procedures and risk characteristics of the financial product information portfolio from the perspective of product compatibility. The rationality verification also includes conducting a market adaptability assessment, combining the current macroeconomic situation, financial market fluctuations, industry development trends and regulatory policy changes to analyze the feasibility and potential returns of the financial product information portfolio in the actual market environment. The rationality verification also includes backtesting analysis of historical data and forecast simulation of future market trends to evaluate the performance of the financial product information portfolio in different market scenarios to ensure that the financial products recommended to small and micro enterprises not only meet the actual needs of the enterprises, but also can operate stably in the market.

[0015] S8. Output and feedback of recommendation results: The financial product information combination verified in the above step S7 is recommended to small and micro enterprise users in an intuitive and easy-to-understand manner, and detailed financial product descriptions are provided. Then, a user feedback mechanism is established. The financial product descriptions include an introduction to the functions of the financial products, the combination advantages, and potential risk warning information. The user feedback mechanism collects the experiences and opinions of small and micro enterprises in the process of using the recommended financial product information combination, which facilitates the subsequent continuous optimization and improvement of the recommendation method to continuously improve the accuracy and practicality of the recommendations.

[0016] Furthermore, in step S1, the credit products include credit loans, mortgage loans and supply chain finance loans, the insurance products include property insurance, credit guarantee insurance and loan guarantee insurance, the payment settlement products include electronic payment platforms and account management systems, the investment and wealth management products include money funds and bond funds, the repayment methods include equal principal and interest, interest first and principal later, and one-time repayment of principal and interest, and the term structure includes short-term, medium-term and long-term.

[0017] Furthermore, in step S3, the lexical semantic analysis adopts a pre-trained language model, which includes BERT and GPT, to perform feature extraction on text data in the financial and ecological fields to obtain more accurate word vector representation, thereby capturing the semantic similarity and semantic association between words. The syntactic dependency analysis uses a deep learning model, which is a dependency syntactic analyzer based on the Transformer architecture. It performs syntactic structure analysis on complex domain sentences, clarifies the dependency relationship between each component in the sentence, and provides a basis for the subsequent semantic role labeling. The semantic role labeling uses an end-to-end model based on a neural network. The end-to-end model based on a neural network is an LSTM-CRF model combined with an attention mechanism. On the basis of the syntactic analysis, the semantic role of each predicate in the sentence is accurately identified. The semantic roles include agent, object, time, and place, further deepening the understanding of the semantic information of domain knowledge.

[0018] Furthermore, in step S3, the machine learning-based knowledge association algorithm is calculated based on a deep learning graph neural network model, and the deep learning-based graph neural network model is a Graph Neural Network. Feature extraction and representation learning are performed on the financial domain knowledge sub-graph and the ecological domain knowledge sub-graph, and the similarity between the two graphs is calculated using a graph matching algorithm to discover potential knowledge associations. Then, based on an association mining algorithm combining rules and statistics, domain-specific rules are first formulated. The domain-specific rules include similarity rules based on the functions of financial products and the survival strategies of ecological species. Then, through statistical analysis methods, the statistical analysis methods include co-occurrence analysis and association rule mining, knowledge association relationships that conform to the rules are mined in large-scale domain data, so as to achieve an intrinsic and effective connection between the two domain knowledge sub-graphs.

[0019] Furthermore, in step S5, the internal rate of return of the profitability indicator is accurately calculated through a cash flow discounting model, taking into account the cash inflows and outflows of the financial product throughout its life cycle. The net present value of the profitability indicator is discounted based on the market interest rate and the expected return of the product to reflect the current value of the product, thereby improving the accuracy of the quantitative evaluation model of the competition and complementary relationship.

[0020] The credit risk of the risk indicators is assessed using a credit scoring model based on big data, taking into account the financial status, credit record, and operational stability of small and micro enterprises. The market risk of the risk indicators is quantified using a value-at-risk (VaR) model or a conditional value-at-risk (CVaR) model to measure the maximum loss that a financial product may face under a certain confidence level.

[0021] The liquidity risk of the liquidity indicator is assessed by calculating the time required for cash conversion and the capital turnover efficiency index. The cash conversion speed of the liquidity indicator is quantified by analyzing factors such as the trading activity and market depth of financial products. The capital turnover cycle of the liquidity indicator is calculated based on the business model of small and micro enterprises and the repayment arrangement of financial products.

[0022] The interest rate cost in the cost indicator is accurately calculated based on the product's interest rate pricing formula, and the handling fee rate in the cost indicator is obtained by summarizing and statistically analyzing various handling fee items.

[0023] Furthermore, in step S5, the personalized demand characteristics of the small and micro enterprises are realized by conducting a detailed demand survey on the small and micro enterprises. The demand survey includes but is not limited to questionnaire surveys, face-to-face interviews and data analysis methods to collect the degree of attention of small and micro enterprises to various attributes of financial products at different development stages, different industry characteristics and different business objectives.

[0024] The method of assigning corresponding weights to different demand dimensions of small and micro enterprises specifically involves using the analytic hierarchy process (AHP) to construct a hierarchical model, decomposing the financial demand goals of small and micro enterprises into multiple levels of criteria and indicators, determining the relative importance of elements at each level through pairwise comparison, and thus calculating the weight vectors of different demand dimensions. The weights are then adjusted and optimized in combination with the objective weighting method of the entropy weight method to ensure that the weights can reflect both the subjective demand preferences of small and micro enterprises and the objective information of the data.

[0025] Furthermore, in step S6, the genetic algorithm (GA) dynamically adjusts the crossover probability and mutation probability parameters, and adopts an adaptive parameter adjustment strategy according to the changes in the fitness function value during the search process. For example, when the fitness of individuals in the population tends to be consistent, the mutation probability is appropriately increased to increase the diversity of the population and avoid the algorithm from falling into the local optimal solution. The particle swarm optimization algorithm (PSO) dynamically adjusts parameters such as the inertia weight and the learning factor according to the changes in the flight speed and position of the particles, so that the particles achieve a balance between global search and local search, thereby improving the search efficiency and accuracy of the algorithm. The simulated annealing algorithm (SA) adaptively adjusts the cooling rate according to the changes in the current temperature and the objective function value to ensure that the algorithm converges to the global optimal solution or an approximate global optimal solution within a reasonable time.

[0026] Furthermore, in step S7, the macroeconomic situation analysis is to select key macroeconomic indicators such as GDP growth rate, inflation rate, and interest rate level, establish an econometric model between macroeconomic indicators and financial product demand and returns, and evaluate the market adaptability of the financial product information combination by analyzing the impact of changes in macroeconomic variables on the financial product market.

[0027] The financial market fluctuations are analyzed by using a time series analysis model combined with a machine learning algorithm to analyze the main indicators of the financial market, comprehensively judge the future trend of the financial market, and provide a basis for adjusting the information combination of the financial product. The time series analysis model adopts the ARIMA model, and the main indicators of the financial market include stock indexes, bond yields and exchange rates. The machine learning algorithm is the support vector machine (SVM) algorithm.

[0028] The industry development trend analysis analyzes the market concentration, competitive landscape, and threat factors of new entrants in the industries where small and micro enterprises are located, and uses Porter's Five Forces Model tool to evaluate the impact of the industry's competitive environment on the demand for financial products, ensuring that the recommended financial product information portfolio can adapt to the dynamic changes in industry development.

[0029] Compared with the prior art, the present invention has the advantages of:

[0030] (1) This solution accurately meets financial needs. In terms of in-depth analysis and graph construction of financial knowledge, it comprehensively sorts out the multi-dimensional characteristics of financial products to construct knowledge sub-graphs, changing the previous limitations of relying solely on historical transaction data and financial statement analysis and recommendations. It can better adapt to the diversified and dynamically changing financial needs of small and micro enterprises and effectively improve the accuracy of recommendations. For example, by carefully analyzing the interest rate pricing, repayment methods and other attributes of various products such as credit, insurance, payment settlement and investment and wealth management products, it can accurately match products to the financing, risk avoidance and fund management needs of enterprises at specific stages.

[0031] (2) This solution enhances the accuracy of cross-domain knowledge application by systematically collecting ecological domain knowledge and presenting it in a structured manner, and then integrating it with advanced natural language processing and machine learning technologies to form a cross-domain knowledge graph. This overcomes the problem of lack of systematicity and depth in the application of cross-domain knowledge in existing methods, and accurately integrates financial and ecological domain knowledge to make the recommendation basis more scientific and comprehensive. For example, based on an in-depth understanding of the relationship between ecological species and financial products, it can explore potential product associations and provide a richer perspective for recommendations.

[0032] (3) This solution optimizes product portfolio and evaluation, and constructs a quantitative evaluation model for the competition and complementary relationship of financial products. It fully considers key indicators such as profitability, risk, liquidity and cost, and empowers small and micro enterprises with personalized demand characteristics. It solves the defect of previous recommendations that did not deeply analyze the complex relationship between products. It uses intelligent optimization algorithms to search for product portfolios, and can dynamically adjust portfolio parameters according to enterprise needs to achieve multi-objective optimization such as maximizing the satisfaction of comprehensive financial needs. It has also been verified by multiple aspects of rationality to ensure that the product portfolio is both in line with the actual situation of the enterprise and can operate stably under different market environments, greatly improving the effectiveness and reliability of recommendations.

[0033] (4) This plan will enhance user experience and continuously improve methods. It will intuitively recommend verified product combinations to small and micro enterprises and provide detailed explanations. At the same time, it will establish a user feedback mechanism to facilitate enterprises to understand the products and help to continuously optimize the recommendation methods. It can continuously improve the accuracy and practicality of recommendations, better serve the financial needs of small and micro enterprises, and promote their healthy development. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 Schematic diagram of the method steps of the present invention.

[0035] Figure 2 This is a mind map of step S1 of the present invention.

[0036] Figure 3 This is a mind map of step S2 of the present invention.

[0037] Figure 4This is a mind map of step S3 of the present invention.

[0038] Figure 5 This is a mind map of step S4 of the present invention.

[0039] Figure 6 This is a mind map of step S5 of the present invention.

[0040] Figure 7 This is a mind map of step S6 of the present invention.

[0041] Figure 8 This is a mind map of step S7 of the present invention.

[0042] Figure 9 This is a mind map of step S8 of the present invention. DETAILED DESCRIPTION

[0043] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the specification of the present invention; it is obvious that the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0044] Example 1

[0045] Combined with the accompanying drawings Figures 1-9 , the financial technology product recommendation process of a small and micro manufacturing enterprise.

[0046] S1. In-depth analysis and map construction of financial knowledge.

[0047] For these small and micro enterprises, we first collected information from the financial sector. Credit products include bank-provided credit loans, with interest rates fluctuating within a certain range based on the small and micro enterprise's credit rating, equal repayments of principal and interest, and a medium-term maturity structure (1-3 years). Mortgage loans, secured by the enterprise's fixed assets, offer relatively low interest rates, with repayments of interest first and principal later, and long-term maturities (3-5 years). Insurance products include property insurance, protecting the enterprise's fixed assets from losses due to natural disasters and other unexpected events; credit guarantee insurance, used to increase the credit limit of the enterprise's loan application; payment and settlement products utilize commonly used electronic payment platforms to facilitate daily transaction settlement; and investment and wealth management products utilize money market funds, which offer high liquidity and low risk.

[0048] Using graph database technology, these financial products and their related attributes are constructed into a knowledge sub-graph in the financial field. Specific credit loan nodes are connected to attribute nodes such as their interest rate range, equal principal and interest repayment method, medium-term term, and applicable size restrictions for small and micro manufacturing enterprises. Edges represent the logical relationship between them, such as the interest rate of credit loans is affected by the credit rating of small and micro enterprises.

[0049] S2. Collection and structured representation of ecological domain knowledge.

[0050] Gathering ecosystem knowledge reveals that within the ecological environment of this manufacturing SME, there are both competitive and cooperative relationships between different species. For example, within the ecological zone surrounding the factory, there are insect populations that feed on factory waste (similar to predation in resource acquisition strategies). These insects have a symbiotic relationship with the surrounding plants, helping them pollinate and providing habitat for the insects. At the same time, there are also carnivorous populations that feed on small mammals, creating a competitive relationship between these species and competing for the same food resources.

[0051] The Web Ontology Language (OWL) technology is used to structure the knowledge of competition and cooperation relationships among these species and construct an ecological knowledge subgraph. For example, the insect population node is connected to the plant population node through a symbiotic relationship edge, and is connected to the carnivore population node through a competitive relationship edge. At the same time, it connects ecological processes such as energy flow (insects obtain energy from plants, and carnivores obtain energy from insects) and related ecological factors (such as habitat type, etc.) nodes.

[0052] S3. Fusion and association of cross-domain knowledge graphs.

[0053] With the help of natural language processing technology, a pre-trained BERT model was used to perform lexical semantic analysis on text data in the financial and ecological fields, and extract word vector representations of text related to financial products and ecological species. For example, the "financing" function of credit loans and the "energy acquisition" of insects to obtain food resources have certain semantic similarities. This association can be preliminarily discovered through the similarity calculation of word vectors.

[0054] A dependency parser based on the Transformer architecture is used to perform syntactic dependency analysis on complex sentences, such as "Small and micro enterprises apply for credit loans for production capital turnover, just like insects looking for food to survive." This clarifies the dependency relationship between the various components in the sentence and provides a basis for semantic role labeling.

[0055] The LSTM-CRF model combined with the attention mechanism is used for semantic role labeling to further understand the semantic information of domain knowledge. On this basis, the deep learning-based Graph Neural Network (GNN) is used to extract features and learn representations of the financial domain knowledge subgraph and the ecological domain knowledge subgraph. The similarity is calculated through a graph matching algorithm to discover potential knowledge associations. At the same time, rules are formulated based on the similarity between the functions of financial products and the survival strategies of ecological species. For example, risk-averse financial products are associated with ecological species with defense mechanisms. Statistical methods such as co-occurrence analysis are then used to explore knowledge associations that meet the rules, and finally integrate them into a cross-domain knowledge graph.

[0056] S4. Analogy mapping between financial needs and ecological resources.

[0057] The small and micro manufacturing enterprises are facing the demand for financing to expand production scale and purchase new equipment, which is similar to the large demand of insect populations in the ecosystem for food resources to maintain population growth. The enterprises also have risk aversion needs, worried that market fluctuations will affect product sales and lead to a break in the capital chain, similar to the demand of certain species in the ecosystem for stable habitats to avoid natural enemies and harsh environments. The demand for capital management is like the demand of species in the ecosystem for the rational allocation and utilization of energy. Various financial products are regarded as different species in the ecosystem. Credit loans are like resources that can provide a large amount of food, which can meet the financing needs of enterprises. Property insurance is like a safe habitat, providing risk aversion protection for enterprises. Money funds are like efficient energy management mechanisms, helping enterprises to reasonably manage idle funds.

[0058] S5. Construction of a quantitative evaluation model for competition and complementary relationships.

[0059] For profitability indicators, the internal rate of return of credit loans is calculated using a cash flow discounting model, taking into account the company's repayment cash flow and interest expenses during the loan period. The net present value is obtained by discounting the current market interest rate and expected returns. For example, after calculation, the internal rate of return of the credit loan applied for by the company is 8%, and the net present value is positive, indicating that it has certain attractiveness in terms of returns.

[0060] In terms of risk indicators, credit risk is assessed using a big data-based credit scoring model. Taking into account the company's financial statements, tax records and other information, the company's credit score is good and its credit risk is at a low level. Market risk is quantified using the VaR model at a 95% confidence level. Analysis shows that under market fluctuations, the maximum probability of loss faced by corporate loans is relatively small.

[0061] Among the liquidity indicators, the time required to realize the funds of a credit loan is the gradual realization of the funds according to the repayment plan within the loan period. The capital turnover efficiency is calculated based on the company's production cycle and repayment cycle. The capital turnover cycle is 1 year. The speed of capital realization is at a medium level through analysis of the trading activity and market depth of similar loans in the financial market.

[0062] Among the cost indicators, the interest cost of credit loans is accurately calculated as 6% annualized based on the bank's interest rate pricing formula, and the handling fee rate is 1% of the loan amount.

[0063] Through questionnaire surveys and data analysis, we understand the personalized demand characteristics of small and micro enterprises. We find that enterprises pay more attention to the timeliness and cost control of capital financing in their current development stage. We use the analytic hierarchy process (AHP) to construct a hierarchical structure model, decomposing the financial demand goals into criteria such as capital financing, risk avoidance, and fund management. Through pairwise comparison, we determine the relative importance of elements at each level, and calculate that the weight of the capital financing demand dimension is 0.5, the weight of the risk avoidance demand dimension is 0.3, and the weight of the fund management demand dimension is 0.2. We adjust and optimize them with the entropy weight method, and finally determine the weight vector. This is used to quantitatively evaluate the competitive and complementary relationship between different financial products in meeting the various financial needs of enterprises.

[0064] S6. Product portfolio search based on optimization algorithm.

[0065] A genetic algorithm (GA) is used to search in a cross-domain knowledge graph with multi-objective functions such as maximizing the satisfaction of the comprehensive financial needs of small and micro enterprises, minimizing financial costs, diversifying risks, and stabilizing returns. During the search process, the crossover probability and mutation probability parameters are dynamically adjusted according to the changes in the fitness function value. For example, the initial crossover probability is set to 0.8 and the mutation probability is 0.1. When the fitness of the individuals in the population tends to be consistent, the mutation probability is increased to 0.3 to increase the diversity of the population and avoid falling into a local optimal solution. After multiple iterations, potential financial product information combinations are generated, such as a combination of a credit loan of 500,000 yuan, a property insurance coverage of 2 million yuan, and a money fund investment of 300,000 yuan, and the weight distribution of each product in the combination is determined, with the credit loan weight being 0.5, the property insurance weight being 0.3, and the money fund weight being 0.2.

[0066] S7. Verification of the rationality of product portfolio.

[0067] From the perspective of product compatibility, the company has met the application conditions for the credit loan, the usage period matches the company's production equipment procurement and repayment plan, the company is proficient in the operating procedures, and the risk characteristics complement property insurance and money funds. There is no conflict or incompatibility.

[0068] A market adaptability assessment was conducted and the current macroeconomic situation was analyzed. The GDP growth rate was stable at 6%, the inflation rate was controlled at 3%, and the interest rate level was moderate. By establishing an econometric model, it was concluded that the financial product information portfolio has good adaptability in the current macroeconomic environment and the expected returns are stable. The ARIMA model combined with the support vector machine (SVM) algorithm was used to analyze the fluctuations in the financial market. The stock index and bond yield fluctuations are within a reasonable range, the exchange rate is stable, and the impact on the portfolio is small. The development trend of the manufacturing industry was analyzed through Porter's Five Forces Model. The market concentration is moderate, the competitive landscape is stable, and the threat of new entrants is small. The portfolio can meet the financial needs of enterprises in the development of the industry. At the same time, a backtest analysis of historical data was conducted to simulate the performance of enterprises using this portfolio in similar market scenarios in the past. The results showed that the financial status and operating stability of the enterprises have been effectively improved. The forecast simulation of future market trends also shows that the portfolio is feasible in the future.

[0069] S8. Output and feedback of recommendation results.

[0070] Recommend a verified financial product information combination to the small and micro enterprise user, and provide detailed financial product descriptions. For example, explain to the enterprise the credit loan amount, interest rate, repayment method and the advantage of quickly solving financing problems, the coverage of property insurance, the claims process and how to reduce corporate risks, the income characteristics, liquidity advantages and management role of money market funds for idle funds. At the same time, establish a user feedback mechanism to collect their experience and opinions during the process of enterprises using the recommended product combination. For example, enterprises can provide feedback on whether the credit loan approval process can be further simplified and whether the income of money market funds can be more stable, so as to continuously optimize and improve the recommendation method in the future.

[0071] Example 2

[0072] Combined with the accompanying drawings Figures 1-9 , the financial technology product recommendation process of a small and micro enterprise with technology services.

[0073] S1. In-depth analysis and map construction of financial knowledge.

[0074] For small and micro-sized technology service enterprises, credit products include special loans for high-tech enterprises with preferential interest rates, a one-time repayment method of principal and interest, and a short-term maturity structure (6 months to 1 year) to meet the short-term funding needs of enterprises' R&D projects. Insurance products include technology insurance to protect enterprises from technical and intellectual property risks during the R&D process. Payment and settlement products use account management systems with cross-border payment functions to facilitate business transactions between enterprises and international customers. Investment and wealth management products are bond funds with relatively stable returns, suitable for enterprises to preserve and increase the value of their funds.

[0075] When constructing a knowledge sub-graph in the financial field, the special loan node for high-tech enterprises is connected to attribute nodes such as preferential interest rates, one-time principal and interest repayment methods, short-term terms, and applicable scale and industry types of small and micro-sized technology service enterprises, clarifying the logical relationship between each product and attribute.

[0076] S2. Collection and structured representation of ecological domain knowledge.

[0077] In the ecological environment where small and micro-sized technology service enterprises are located, there is a competitive relationship between innovative enterprises and traditional enterprises, similar to the competition between different species in an ecosystem for ecological niches. For example, in the same science and technology park, small and micro-sized enterprises compete with other similar enterprises for limited talent resources and market share. At the same time, they have cooperative relationships with upstream and downstream enterprises, such as cooperating with scientific research institutions to obtain technical support (similar to a symbiotic relationship) and cooperating with suppliers to ensure the supply of raw materials (similar to the predator-prey relationship in resource acquisition strategies).

[0078] OWL technology is used to construct a knowledge sub-graph in the ecological field, with small and micro-sized technology service enterprises and their competing and cooperating enterprises as nodes. Different types of edges are used to represent relationships such as competition and cooperation, and to connect related ecological processes and ecological factor nodes, such as ecological process nodes such as talent flow and technological innovation.

[0079] S3. Fusion and association of cross-domain knowledge graphs.

[0080] The BERT model is used for lexical semantic analysis to explore the semantic associations between texts in the fields of science and technology finance and ecology. For example, the "risk protection" function of science and technology insurance has semantic similarities with the "self-protection" mechanism of species in the ecosystem. The deep learning dependency parser is used to parse complex sentences, such as "Small and micro-sized science and technology service enterprises purchase science and technology insurance to resist R&D risks, just like species develop defense mechanisms to cope with environmental changes in biological evolution." The sentence structure relationship is determined, and semantic role labeling is performed through an end-to-end neural network model to further understand the semantic information. Based on the deep learning graph matching algorithm and the association mining algorithm combining rules and statistics, the knowledge sub-graph of the financial field and the knowledge sub-graph of the ecological field are integrated, a mapping relationship is established, and a cross-domain knowledge graph is formed. For example, it is found that science and technology insurance is associated with ecological species with defense strategies. The co-occurrence of science and technology enterprises and related ecological factors is statistically analyzed to strengthen this knowledge association.

[0081] S4. Analogy mapping between financial needs and ecological resources.

[0082] The financing needs of these small and micro-sized technology service enterprises are mainly used for R&D investment and market expansion, similar to the needs of species in an ecosystem for key resources in order to evolve and expand their populations. The risk aversion needs stem from the uncertainty of technological research and development and the fierce market competition, just like the needs of species in an ecosystem for a safe environment when facing environmental changes and threats from natural enemies. The fund management needs are to improve the efficiency of fund use, similar to the needs of species in an ecosystem for efficient energy utilization.

[0083] Comparing financial products to ecological resources, special loans for high-tech enterprises are key resources for R&D financing, technology insurance is a safety guarantee for dealing with risks, and bond funds are an effective tool for optimizing fund management.

[0084] S5. Construction of a quantitative evaluation model for competition and complementary relationships.

[0085] In terms of profitability indicators, the internal rate of return of the bond fund is calculated to be 7% through the cash flow discounting model based on the returns of its investment portfolio and market interest rate fluctuations. The net present value is calculated based on the expected returns and market interest rate discounts, reflecting its value in the current market environment.

[0086] Among the risk indicators, the risk assessment of technology insurance comprehensively considers factors such as the company's technological research and development risks and intellectual property risks, and quantifies them through professional risk assessment models. Credit risk is based on the company's credit record and financial status, and is assessed using a big data credit scoring model. The company has good credit and the risk is within a controllable range. Market risk is analyzed using the CVaR model at a 90% confidence level to determine the potential loss risk faced by the company during market fluctuations.

[0087] Among the liquidity indicators, the bond fund's cash conversion rate is analyzed based on its trading activity and market depth, and has a certain degree of liquidity. The capital turnover cycle is calculated based on the company's capital utilization plan and the fund's redemption regulations, which is 1.5 years.

[0088] Among the cost indicators, the interest rate cost of special loans for high-tech enterprises is determined to be 5% annualized based on government support policies and bank pricing formulas, and the handling fee rate is relatively low. Through face-to-face interviews and data analysis, we understand the personalized demand characteristics of enterprises. At the current stage, enterprises are more concerned about risk aversion and capital management. The hierarchical structure model is constructed using the analytic hierarchy process (AHP), and the weight of the risk aversion demand dimension is determined to be 0.4, the weight of the capital management demand dimension is 0.4, and the weight of the financing demand dimension is 0.2. The entropy weight method is used for adjustment and optimization, and a quantitative evaluation of the competition and complementary relationship of financial products is conducted.

[0089] S6. Product portfolio search based on optimization algorithm.

[0090] Using the particle swarm optimization (PSO) algorithm, a search is conducted in a cross-domain knowledge graph with multi-objective functions such as maximizing the satisfaction of the company's comprehensive financial needs. Parameters such as the inertia weight and learning factor are dynamically adjusted according to the particle's flight speed and position changes. For example, the initial inertia weight is 0.9, and the learning factors are 1.5 and 1.6, respectively. As the search progresses, when it is found that particles are excessively clustered in local areas, the inertia weight is appropriately reduced to 0.7 and the learning factors are increased to 2 and 2, so that the particles achieve a balance between global and local searches. After algorithm iteration, a potential financial product portfolio is generated, such as an 800,000 yuan special loan for high-tech enterprises, a 3 million yuan technology insurance coverage, and a 500,000 yuan bond fund investment. The weight distribution of each product in the portfolio is also determined, with the special loan weight being 0.4, the technology insurance weight being 0.4, and the bond fund weight being 0.2.

[0091] S7. Verification of the rationality of product portfolio.

[0092] In terms of product compatibility, enterprises meet the application requirements for special loans, the loan period aligns with the R&D project cycle, the operational process is convenient, and the combination with technology insurance and bond funds provides good risk diversification. In the market adaptability assessment, an analysis of the macroeconomic situation revealed a high GDP growth rate, policy support for the science and technology innovation industry, and favorable interest rates for corporate financing. This combination has advantages in this macroeconomic environment. Analysis of financial market fluctuations using time series analysis models and machine learning algorithms revealed a stable technology finance market with minimal impact on the portfolio. Using Porter's Five Forces Model to analyze the development trends of the technology services industry, the industry is growing rapidly, facing fierce competition but numerous opportunities for innovation. This combination can meet the financial needs of enterprises in the industry's development. Backtesting of historical data and simulations of future market trend forecasts show that enterprises using this combination have maintained a good financial position during past market fluctuations and are expected to achieve stable development in the future.

[0093] S8. Output and feedback of recommendation results.

[0094] Recommend the verified financial product portfolio to small and micro-sized technology service enterprises, provide detailed product descriptions, including preferential policies for special loans, coverage and claims processing of technology insurance, investment strategies and expected returns of bond funds, etc. Establish a user feedback mechanism to collect opinions from enterprises during use, such as whether the claims processing speed of technology insurance can be improved, investment recommendations for bond funds, and other feedback information, so as to optimize and improve the recommendation method.

[0095] The above examples more clearly demonstrate the specific application process and effectiveness of the intelligent recommendation method for fintech products for small and micro enterprises across different types of small and micro enterprises, further validating the feasibility and effectiveness of the method. In actual application, the method can be flexibly adjusted and optimized based on the specific circumstances of different small and micro enterprises to meet their diverse financial needs.

[0096] The above description is merely a preferred embodiment of the present invention; however, the scope of protection of the present invention is not limited thereto. Any person skilled in the art who, within the technical scope disclosed by the present invention, makes equivalent substitutions or modifications based on the technical solutions and improved concepts of the present invention shall be covered by the scope of protection of the present invention.

Claims

1. A method for intelligently recommending financial technology products for small and micro enterprises, characterized in that: It includes the following interrelated and sequential steps: S1. In-depth analysis and graph construction of financial domain knowledge: Targeting the financial sector, systematically collect information on various financial products, including credit products, insurance products, payment and settlement products, and investment and wealth management products. For each type of financial product, we analyze in detail its interest rate pricing mechanism, repayment method, term structure, applicable small and micro enterprise industry types, enterprise size restrictions, risk rating system, and multi-dimensional characteristics of regulatory compliance requirements. We use graph database technology to construct a financial domain knowledge subgraph. The nodes of the financial domain knowledge subgraph represent various financial products and their related attributes, and the edges of the financial domain knowledge subgraph represent the logical relationships between products and attributes, as well as between different products. S2. Collection and structured representation of ecological domain knowledge: Systematically collect knowledge about species competition and cooperation in the ecosystem domain, including the distribution of different species in ecological niches, resource acquisition strategies, energy flow and material circulation patterns in ecosystems, and population dynamics. Resource acquisition strategies include predation, symbiosis, and parasitism. Utilize the web ontology language (OWL) to structure this knowledge about species competition and cooperation, and construct an ecological domain knowledge subgraph. The nodes of the ecological domain knowledge subgraph correspond to different species, ecological processes, and related ecological factors, and the edges of the ecological domain knowledge subgraph represent the interactions between species and the associations between ecological processes and various factors. S3. Fusion and association of cross-domain knowledge graphs: Using semantic analysis methods in natural language processing technology, including lexical semantic analysis, syntactic dependency analysis, and semantic role labeling, we gain a deep understanding of the semantic connotations of knowledge in the financial and ecological fields. On this basis, we use a machine learning-based knowledge association algorithm to establish a mapping relationship between the financial knowledge sub-graph and the ecological knowledge sub-graph, integrating them to form a cross-domain knowledge graph. This cross-domain knowledge graph not only intuitively reflects the inherent connection between the knowledge in the two fields, but also provides a solid data foundation for subsequent knowledge migration and fusion. S4. Analogy Mapping of Financial Needs and Ecological Resources: Compare the financial needs of small and micro enterprises to the resource needs of species in an ecosystem. Consider each type of financial product information as a different species in the ecosystem. Each type of financial product information meets the financial needs of small and micro enterprises through its unique multi-dimensional characteristics, just as species obtain the required resources through their own survival strategies. Financial needs include financing needs, risk aversion needs, and fund management needs, while resource needs include food, habitat, and sunlight. S5. Construction of a Quantitative Assessment Model for Competition and Complementarity: Based on the theories and mechanisms of competition and cooperation among species in ecosystems, a quantitative assessment model for the competition and complementarity of financial product information is constructed. This quantitative assessment model also fully considers multiple key attribute indicators of the financial product information to improve the accuracy of the construction of the quantitative assessment model for competition and complementarity. These key attribute indicators include profitability, risk, liquidity, and cost indicators. Furthermore, this quantitative assessment model for competition and complementarity of financial product information incorporates the personalized needs of small and micro enterprises. Using the multi-criteria decision analysis method of the Analytic Hierarchy Process (AHP), it assigns corresponding weights to the different demand dimensions of small and micro enterprises, and quantitatively assesses the competition and complementarity of different financial product information in meeting the various financial needs of small and micro enterprises. S6. Product portfolio search based on optimization algorithms: Using an intelligent optimization algorithm with the multi-objective function of maximizing the satisfaction of the comprehensive financial needs of small and micro enterprises, minimizing financial costs, risk diversification, and return stability, a search is conducted in the cross-domain knowledge graph to generate potential financial product information combinations based on the cross-domain knowledge graph. The intelligent optimization algorithm includes a genetic algorithm (GA), a particle swarm optimization algorithm (PSO), and a simulated annealing algorithm (SA). During the search process, the intelligent optimization algorithm dynamically adjusts the constituent parameters of the financial product information combination, including the product type, product quantity, and the weight distribution of each product in the combination, to adapt to the dynamically changing financial needs of different small and micro enterprises at different development stages and market environments; S7. Product Portfolio Rationality Verification: Conduct a comprehensive rationality verification of the generated financial product portfolio. This includes carefully examining, from the perspective of product compatibility, whether there are any conflicts or incompatibilities in the application requirements, usage period, operating procedures, and risk characteristics of the financial product portfolio. This rationality verification also includes conducting a market adaptability assessment, analyzing the feasibility and potential returns of the financial product portfolio in the real market environment based on the current macroeconomic situation, financial market fluctuations, industry development trends, and regulatory policy changes. This rationality verification also includes backtesting analysis of historical data and forecast simulation of future market trends to evaluate the performance of the financial product portfolio in different market scenarios, ensuring that the financial products recommended to small and micro enterprises not only meet the actual needs of the enterprises but also operate stably in the market. S8. Output and feedback of recommendation results: The financial product information combination verified in the above step S7 is recommended to small and micro enterprise users in an intuitive and easy-to-understand manner, and detailed financial product descriptions are provided. Then, a user feedback mechanism is established. The financial product descriptions include an introduction to the functions of the financial products, the combination advantages, and potential risk warning information. The user feedback mechanism collects the experiences and opinions of small and micro enterprises in the process of using the recommended financial product information combination, which facilitates the subsequent continuous optimization and improvement of the recommendation method to continuously improve the accuracy and practicality of the recommendations.

2. The intelligent recommendation method for fintech products for small and micro enterprises according to claim 1, characterized in that: In step S1, the credit products include credit loans, mortgage loans and supply chain finance loans, the insurance products include property insurance, credit guarantee insurance and loan guarantee insurance, the payment settlement products include electronic payment platforms and account management systems, the investment and financial products include money funds and bond funds, the repayment methods include equal principal and interest, interest first and principal later, and one-time principal and interest repayment, and the term structure includes short-term, medium-term and long-term.

3. The intelligent recommendation method for fintech products for small and micro enterprises according to claim 1, characterized in that: In step S3, the lexical semantic analysis adopts a pre-trained language model, which includes BERT and GPT, to perform feature extraction on text data in the financial and ecological fields to obtain more accurate word vector representation, thereby capturing the semantic similarity and semantic association between words. The syntactic dependency analysis uses a deep learning model, which is a dependency syntactic analyzer based on the Transformer architecture. It performs syntactic structure analysis on complex domain sentences, clarifies the dependency relationship between each component in the sentence, and provides a basis for the subsequent semantic role labeling. The semantic role labeling uses an end-to-end model based on a neural network. The end-to-end model based on the neural network is an LSTM-CRF model combined with an attention mechanism. Based on the syntactic analysis, the semantic role of each predicate in the sentence is accurately identified. The semantic roles include agent, object, time, and place, further deepening the understanding of the semantic information of domain knowledge.

4. The intelligent recommendation method for fintech products for small and micro enterprises according to claim 1, characterized in that: In step S3, the knowledge association algorithm based on machine learning is calculated based on a graph neural network model based on deep learning. The graph neural network model based on deep learning is GraphNeuralNetwork. Feature extraction and representation learning are performed on the financial domain knowledge sub-graph and the ecological domain knowledge sub-graph, and the similarity between the two graphs is calculated using a graph matching algorithm to discover potential knowledge associations. Then, domain rules are first formulated based on an association mining algorithm that combines rules and statistics. The domain rules include similarity rules based on the functions of financial products and the survival strategies of ecological species. Then, through statistical analysis methods, the statistical analysis methods include co-occurrence analysis and association rule mining, knowledge association relationships that conform to the rules are mined in large-scale domain data, so as to achieve an intrinsic and effective connection between the two domain knowledge sub-graphs.

5. The intelligent recommendation method for fintech products for small and micro enterprises according to claim 1, characterized in that: In step S5, the internal rate of return of the profitability indicator is accurately calculated using a discounted cash flow model, taking into account the cash inflows and outflows of the financial product throughout its life cycle. The net present value of the profitability indicator is discounted based on the market interest rate and the expected return of the product to reflect the current value of the product, thereby improving the accuracy of the quantitative evaluation model of the competition and complementary relationship. The credit risk of the risk indicator is assessed using a big data-based credit scoring model, which comprehensively considers the financial status, credit history, and operational stability of small and micro enterprises. The market risk of the risk indicator is quantified using a VaR model or a conditional VaR model to measure the maximum loss that a financial product may face under a certain confidence level. The liquidity risk of the liquidity indicator is assessed by calculating the time required to realize funds and the fund turnover efficiency index. The fund realization speed of the liquidity indicator is quantified by analyzing factors such as the trading activity and market depth of financial products. The fund turnover period of the liquidity indicator is calculated based on the business model of the small and micro enterprise and the repayment schedule of the financial product. The interest rate cost in the cost indicator is accurately calculated based on the product's interest rate pricing formula, and the handling fee rate in the cost indicator is obtained by summarizing and statistically analyzing various handling fee items.

6. The intelligent recommendation method for fintech products for small and micro enterprises according to claim 1, characterized in that: In step S5, the personalized demand characteristics of the small and micro enterprises are realized by conducting a detailed demand survey on the small and micro enterprises. The demand survey includes but is not limited to questionnaire surveys, face-to-face interviews, and data analysis to collect the attention paid by small and micro enterprises to various attributes of financial products at different development stages, different industry characteristics, and different business objectives; The method of assigning corresponding weights to different demand dimensions of small and micro enterprises specifically involves using the hierarchical analysis method (AHP) to construct a hierarchical structure model, decomposing the financial demand goals of small and micro enterprises into multiple levels of criteria and indicators, determining the relative importance of elements at each level through pairwise comparison, and thus calculating the weight vectors of different demand dimensions, and adjusting and optimizing the weights in combination with the objective weighting method of the entropy weight method to ensure that the weights can reflect both the subjective demand preferences of small and micro enterprises and the objective information of the data.

7. The intelligent recommendation method for fintech products for small and micro enterprises according to claim 1, characterized in that: In step S6, the genetic algorithm GA dynamically adjusts the crossover probability and mutation probability parameters, and adopts an adaptive parameter adjustment strategy according to the change of the fitness function value during the search process. For example, when the fitness of the individuals in the population tends to be consistent, the mutation probability is appropriately increased to increase the diversity of the population and avoid the algorithm from falling into the local optimal solution. The particle swarm optimization algorithm PSO dynamically adjusts parameters such as the inertia weight and the learning factor according to the flight speed and position changes of the particles, so that the particles achieve a balance between global search and local search, thereby improving the search efficiency and accuracy of the algorithm. The simulated annealing algorithm SA adaptively adjusts the cooling rate according to the changes in the current temperature and the objective function value to ensure that the algorithm converges to the global optimal solution or an approximate global optimal solution within a reasonable time.

8. The intelligent recommendation method for fintech products for small and micro enterprises according to claim 1, characterized in that: In step S7, the macroeconomic situation analysis is to select key macroeconomic indicators such as GDP growth rate, inflation rate, and interest rate level, establish an econometric model between macroeconomic indicators and financial product demand and returns, and evaluate the market adaptability of the financial product information portfolio by analyzing the impact of changes in macroeconomic variables on the financial product market; The financial market fluctuations are analyzed by using a time series analysis model combined with a machine learning algorithm to analyze the main indicators of the financial market, comprehensively judge the future trend of the financial market, and provide a basis for adjusting the financial product information portfolio. The time series analysis model adopts the ARIMA model, and the main indicators of the financial market include stock indexes, bond yields, and exchange rates. The machine learning algorithm is the support vector machine (SVM) algorithm; The industry development trend analysis analyzes the market concentration, competitive landscape, and threat factors of new entrants in the industries where small and micro enterprises are located, and uses Porter's Five Forces Model tool to evaluate the impact of the industry's competitive environment on the demand for financial products, ensuring that the recommended financial product information portfolio can adapt to the dynamic changes in industry development.

Images