Financial investment and financing management platform based on artificial intelligence big data

Abstract

The invention discloses a financial investment and financing management platform based on artificial intelligence big data, relates to the technical field of financial data management, and solves the problems that a full-database engine search mode is adopted, not only is computing power resources of a search engine excessively relied on, but also data output is slow due to synchronization. According to the invention, the multi-thread entry regulation and control end uses the preset thread to carry out entry according to the data type, so that the entry efficiency is improved; by monitoring the entry rate and the standard entry rate in real time and dynamically deploying computing power resources, thread load balance is ensured, threads are prevented from scrambling for a memory and losing packets, and the entry effect and data integrity are ensured; an index vector construction end generates a feature vector circle and a node index vector according to a storage node path, and generates an index table in combination with a classification data header, so that rapid data indexing is realized, a complex hierarchy path is converted into a visual index vector, and the indexing accuracy and speed are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of financial data management, and in particular to a financial investment and financing management platform based on artificial intelligence big data. Background Art

[0002] Against the backdrop of rapid development of the financial industry and the deep integration of technology, it is crucial to build a financial investment and financing management platform based on artificial intelligence big data; relying on advanced artificial intelligence algorithms and massive financial data, the platform provides comprehensive, efficient and intelligent services to financial institutions, investors and financiers, aiming to break down information barriers, improve the efficiency of financial resource allocation, enhance risk management capabilities, and promote innovative development of the financial market.

[0003] The application with publication number CN110210962A discloses a one-stop investment and financing information service platform for small and medium-sized enterprises, which adopts a B / S model and includes a WEB application server, a load balancing server, an investment and financing information service system, a database system, a management terminal, a user terminal, a network firewall, and a communication network; the investment and financing information service system is used to provide one-stop investment and financing information services to small and medium-sized enterprises, including a user-end system and a management-end system. The user-end system is used to provide debt financing, equity financing, financial intermediation, policy reporting, corporate credit investigation and other related services to small and medium-sized enterprises, and to establish a reliable and fast investment and financing channel between small and medium-sized enterprises and financial and investment institutions; the management-end system is used to provide online user identity management, debt and equity investment and financing management, business review management, intermediary information management and other related management services to the manager of the investment and financing information service system.

[0004] In the process of data management, the financial data associated with its financial investment and financing are generally stored in a designated storage interval based on the entry time of the corresponding financial data, and then the relevant financial data is indexed based on the header of the corresponding data. However, this type of indexing method has a slow indexing rate due to the huge volume of entered financial data. The full database engine search method is not only overly dependent on the search engine's computing resources, but synchronization will also cause slow data output. The indexing method in the original data management process needs to be improved. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides a financial investment and financing management platform based on artificial intelligence big data, which solves the problem of using a full database engine search method, which not only over-relies on the search engine's computing resources, but also causes slow data output due to synchronization.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a financial investment and financing management platform based on artificial intelligence big data, comprising: The financial data acquisition end collects financial data generated by different data sources within a specified period, where the specified period is a preset period; The multi-threaded input control terminal inputs the financial data collected in this batch based on the preset multiple threads, inputs the financial data of the specified data type according to the inputtable type associated with the corresponding thread, and adjusts the computing power resources associated with a single thread based on the input characteristics generated in real time, so that multiple threads are in a load-balanced state when inputting synchronously. The specific method is as follows: Based on the financial data collected in this batch, classified data of different data types are identified, and each group of classified data corresponds to one data type; According to the preset multiple threads and the enterable types recorded by the corresponding threads, the enterable types are preset types, corresponding to the data types, matching the classification data consistent with the corresponding thread types, and recording the classification data; Record the input rate associated with each thread in real time and mark it as V i , where i represents different threads, and the preset standard input rate of each thread is calibrated as B i , where the standard input rate is the preset rate, and real-time confirmation is available: (B i -V i )>Y1, if it exists, then this thread is calibrated as the thread to be adjusted, where Y1 is the preset fluctuation value; if it does not exist, then no calibration is performed; Based on the calibrated threads to be adjusted and other actually running threads, confirm the computing resources associated with each thread, and allocate the computing resources associated with other threads to the threads to be adjusted. Each time the computing resources associated with the other thread allocation process are one unit, the unit is the preset unit, that is, 1FLOPS. Confirm the thread load balancing associated with each allocation process: Input rate V based on real-time monitoring i and standard input rate B i , confirm the input feature ratio associated with the corresponding thread: TB i =V i ÷B i , and compare the multiple sets of input features associated with multiple threads to TB i Perform variance processing, confirm the standard variance, and confirm whether the standard variance meets the following conditions: standard variance ≤ Y2, where Y2 is the preset value. If so, the allocation process of computing resources is terminated. If not, the allocation is continued and the threads to be adjusted are confirmed in real time until multiple threads are load balanced. The cloud database stores the financial data entered in the current batch, where each different classification data in the financial data is associated with a different storage node path, and the financial data entered in different batches are stored in different classification intervals; The index vector construction end confirms the storage node path of each different classification data in the current batch, and generates a corresponding number of hierarchy circles according to the hierarchical classification of multiple groups of different storage node paths. Then, according to the total number of storage node paths, the hierarchy circles of the specified level are confirmed for the average points. Then, the multiple groups of average points associated with a single group of classification data are locked to the node index vector associated with the corresponding classification data. The specific method is as follows: Based on the storage node path of each different classified data, the first group of hierarchical paths is used as the initial path, and the other node paths that appear successively are confirmed in turn. With the initial path as the center of the circle, a hierarchical circle associated with the second group of hierarchical paths is generated, and based on the multiple groups of storage node paths included in the second group of hierarchical paths, the total number G of different storage node paths is identified from the multiple groups of storage node paths, and then the number of evenly divided points of the hierarchical circle associated with the second group of hierarchical paths is confirmed, and the number of evenly divided points is also G, and the storage node path associated with each different evenly divided point is recorded; Generate the hierarchical circles associated with the third hierarchical path, the latter hierarchical circles are located outside the former hierarchical circles, and the same determination method as the hierarchical circles of the second hierarchical path is used to confirm the equal division points on the circumference of the third hierarchical circle; And so on, until the level circles associated with the last set of level paths and the internal equal division points are confirmed, the generated multiple sets of level circles are recorded as the characteristic vector circles associated with this batch of financial data; According to the storage node path of the classification data associated with the corresponding classification data, the associated equal-division points are selected on the corresponding level circle in turn, and the line segments are connected in turn from the center of the feature vector circle outward to generate the node index vector belonging to this classification data. The node index vectors of all the classification data in this batch are confirmed in the feature vector circle, and the confirmed classification data headers are matched with the corresponding node index vectors one by one to generate an index table, and the generated index table and the feature vector circle associated with the current batch are transmitted to the storage unit for storage; The index output end receives the index instruction input from the outside, and confirms the associated node index vector from the storage unit based on the index instruction, and then directly determines the storage location of the corresponding classification data from the cloud database based on the associated storage node path, and directly indexes and outputs it. The specific method is as follows: Based on the received index instruction, confirm the classification data header associated with the index instruction in the storage unit index table, and then confirm the node index vector associated with the classification data header; Based on the node index vector and the associated characteristic vector circle, the storage location of the classification data is quickly locked. If a single index instruction corresponds to multiple different classification data headers, the node index vector associated with each different classification data header is confirmed, and the storage location of the classification data is quickly locked based on the node index vector and the associated characteristic vector circle; If there is only a single item of classified data, the index output is performed directly; If there are multiple different classification data, the byte quantity associated with the different classification data is determined based on the attributes of the corresponding classification data and marked as Z k , where k represents different classification data and is based on the byte size Z k Sort multiple different classification data from small to large, and index and output the sorted classification data in turn.

[0007] The present invention provides a financial investment and financing management platform based on artificial intelligence big data. Compared with the prior art, it has the following beneficial effects: The present invention uses a multi-threaded input control terminal to use preset threads to input data according to data types, thereby improving input efficiency; by real-time monitoring of the input rate and the standard input rate, dynamically allocating computing resources, ensuring thread load balancing, avoiding thread competition for memory and packet loss, and ensuring input effect and data integrity; The index vector construction end generates a feature vector circle and a node index vector according to the storage node path, and generates an index table in combination with the classification data header to achieve fast data indexing, converting complex hierarchical paths into intuitive index vectors, thus improving index accuracy and speed; The index output terminal receives the index instruction and quickly locates the data storage location based on the index table and the feature vector circle. For multiple data indexes, they are sorted and output in sequence by byte quantity, which not only guarantees the efficiency of index output, but also ensures the orderly output of data, avoids confusion, optimizes the user search experience, and improves the value of data use. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 It is a schematic diagram of the principle framework of the present invention; Figure 2 It is a schematic diagram of determining the characteristic vector circle of the present invention. DETAILED DESCRIPTION

[0009] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0010] First embodiment: Please refer to Figure 1 , the present application provides a financial investment and financing management platform based on artificial intelligence big data, including a financial data acquisition end, a multi-threaded input and control end, a cloud database, an index vector construction end, a storage unit and an index output end, wherein the financial data acquisition end, the multi-threaded input and control end and the cloud database are electrically connected from the output node to the input node in sequence, and the cloud database is electrically connected to the index vector construction end and the index output end input node respectively, and the index vector construction end, the storage unit and the index output end are electrically connected from the output node to the input node in sequence; Among them, the financial data acquisition end collects the financial data generated by different data sources within a specified period, and enters the financial data collected in this batch into the multi-threaded input control end, where the specified period is a preset period, generally 24 hours, that is, the data is collected once a day. Its artificial intelligence needs a large amount of data for training during the simulation training process, so the financial data generated by different sources each time needs to be collected in batches. Its different data sources include but are not limited to financial investment and financing data such as stock exchanges, banking systems, corporate financial reports, etc., and the volume of financial data collected in each batch is huge, and there are a large number of related financial data of different data types, and its data format includes but is not limited to PDF format, text format, audio format, video file, etc.; Among them, the multi-threaded input control end inputs the financial data collected in this batch based on the preset multiple threads, inputs the financial data of the specified data type according to the inputtable type associated with the corresponding thread, and adjusts the computing power resources associated with a single thread based on the input characteristics generated in real time, so that multiple threads are in a load-balanced state when inputting synchronously. The specific method of adjusting the computing power resources in real time is: Based on the financial data collected in this batch, classified data of different data types are identified, and each group of classified data corresponds to one data type; According to the preset multiple threads and the enterable types recorded by the corresponding threads, where the enterable type is a preset type, corresponding to the data type, matching the classification data (whose type is the enterable type and the data type) consistent with the corresponding thread type, and recording the classification data; Record the input rate associated with each thread in real time and mark it as V i , where i represents different threads, and the preset standard input rate of each thread is calibrated as B i , where the standard input rate is a preset rate, which is prepared in advance by relevant operators based on experience, and is confirmed in real time whether: (B i -V i)>Y1, if it exists, then this thread will be calibrated as the thread to be adjusted, where Y1 is the preset fluctuation value, which is prepared by relevant personnel in advance based on experience. If it does not exist, no calibration will be performed; Based on the calibrated threads to be adjusted and other actually running threads, confirm the computing power resources associated with each thread (the computing power resources have been preset in advance and can be understood as computing power values), and allocate the computing power resources associated with other threads to the threads to be adjusted. Each time the computing power resources associated with the other thread allocation process are one unit, the unit is the preset unit, that is, 1FLOPS, and confirm the thread load balancing associated with each allocation process: Input rate V based on real-time monitoring i and standard input rate B i , confirm the input feature ratio associated with the corresponding thread: TB i =V i ÷B i , and compare the multiple sets of input features associated with multiple threads to TB i Perform variance processing, confirm the standard variance, and confirm whether the standard variance meets the following conditions: standard variance ≤ Y2, where Y2 is a preset value, and its specific value is determined by the operator based on experience. If it meets the conditions, the allocation process of computing resources is terminated. If it does not meet the conditions, the allocation continues and the threads to be adjusted are confirmed in real time until multiple threads are load balanced. For example, there are multiple groups of threads. There are only four types of classified data in the financial data. Therefore, the corresponding classified data is entered through the corresponding threads. During the entry process, the real-time entry rate V associated with each thread is confirmed in real time. i , and then based on the preset relevant standards, evaluate whether the corresponding thread is in a normal recording state. When it is in a normal recording state, no calibration is required. When it is not in a normal recording state, computing resources are allocated. A group of units of computing resources are allocated from other threads each time. In the actual process of allocating computing resources in sequence, confirm whether the rate share of each thread is balanced enough. When the computing resources associated with multiple threads are synchronously in a balanced state, it can be effectively guaranteed that multiple threads are in the same load state, so there will be no situation where threads compete for memory, ensuring the recording effect during recording, and no packet loss.

[0011] Among them, the cloud database stores the financial data entered in the current batch, wherein each different classification data in the financial data is associated with a different storage node path. The financial data entered in different batches are stored in different classification intervals. In order to achieve the corresponding time distinction effect, the financial data of different time periods are stored separately. The system will store different classification data in the specified storage node path according to the preset storage program. For example, when downloading data from a specified website or port, the computer will select a group of paths for data storage. Such self-selected paths are the paths associated with the classification data. Among them, the index vector construction end confirms the storage node path of each different classification data in the current batch, and generates a corresponding number of hierarchical circles according to the hierarchical classification of multiple groups of different storage node paths, and then confirms the average points of the hierarchical circles of the specified level according to the total number of storage node paths, and then locks the node index vector associated with the corresponding classification data for multiple groups of average points associated with a single group of classification data, so as to facilitate subsequent rapid indexing. The storage node path associated with a group of classification data is proposed as: 2024.02.19 / users / financial_archive / derivative_data / 202502_derivative_batch_2 / option_data / 20250219_1120, which is just an example here, where 2024.02.19 is the time interval. In order to ensure that each different batch of financial data is stored in a different classification interval, the initial path of each classification data associated with this batch is 2024.02.19, and "2024.02.19" here is the name of the corresponding initial storage path; Based on the storage node path of each different classified data, take the first group of hierarchical paths as the initial path, confirm the other node paths that appear in succession in turn, take the initial path as the center of the circle, generate a hierarchical circle associated with the second group of hierarchical paths, and based on the multiple groups of storage node paths included in the second group of hierarchical paths, identify the total number G of different storage node paths from the multiple groups of storage node paths, and then confirm the number of evenly divided points of the hierarchical circle associated with the second group of hierarchical paths, the number of evenly divided points is also G, and record the storage node path associated with each different evenly divided point (here it can be randomly assigned to ensure that each evenly divided point corresponds to the storage node path associated with the second group of hierarchical paths one by one); Generate the hierarchical circles associated with the third hierarchical path, the latter hierarchical circles are located outside the former hierarchical circles (the radius of the hierarchical circles does not need to be limited, as long as the equal-division points can be determined), and use the same determination method as the hierarchical circles of the second hierarchical path to confirm the equal-division points on the circumference of the third hierarchical circles; And so on, until the level circles associated with the last set of level paths and the internal equal division points are confirmed, the generated multiple sets of level circles are recorded as the characteristic vector circles associated with this batch of financial data; According to the storage node path of the classification data associated with the corresponding classification data, the associated equal-division points are selected on the corresponding level circle in turn, and the line segments are connected in turn from the center of the feature vector circle outward to generate the node index vector belonging to this classification data. The node index vectors of all classification data in this batch are confirmed in the feature vector circle, and the confirmed classification data header (that is, the data title) is matched with the corresponding node index vector one by one to generate an index table, and the generated index table and the feature vector circle associated with the current batch are transmitted to the storage unit for storage; Combination Figure 2 :It is proposed that there are nine groups of currently associated classified data. The storage node paths associated with each group of classified data are divided into five levels in sequence. The first level is the same, which are all storage intervals associated with the corresponding batch. There are six groups of different storage paths in the second level (Q1, Q2, ..., Q6). Therefore, six groups of evenly divided points are confirmed in the first group of level circles, and then the third level is confirmed. There are three groups of different storage paths in the third level, so three groups of evenly divided points are confirmed in the second group of level circles (QQ1, QQ2, QQ3). After confirmation, the evenly divided points of the third group of level circles are 4 groups ( QQQ1, ..., QQQ4), the fourth group of level circles has 9 equal points (QQQQ1, QQQQ2, ..., QQQQ9). Because each different equal point is associated with a different storage node path, the node index vector is confirmed in the feature vector circle according to the storage node path associated with a single group of classified data. The storage node path is the classified data of 2024.02.19 / users / financial_archive / derivative_data / 202502_derivative_batch_2. Figure 2 The node vector path confirmed in is: center of circle-Q1-QQ2-QQQ2-QQQQ2, so the node index vector associated with each different classification data is different.

[0012] Second embodiment: In the specific implementation process of this embodiment, compared with the above embodiment, when indexing, this embodiment has multiple synchronous output processes of index data, and the index output is performed by the relevant index output terminal; Among them, the index output end receives the index instruction input from the outside, and confirms the associated node index vector from the storage unit based on the index instruction, and then directly determines the storage location of the corresponding classification data from the cloud database based on the associated storage node path, and directly indexes and outputs it. The specific operation is as follows: Based on the received index instruction, confirm the classification data header associated with the index instruction in the storage unit index table, and then confirm the node index vector associated with the classification data header; Based on the node index vector and the associated characteristic vector circle, the storage location of the classification data is quickly locked. If a single index instruction corresponds to multiple different classification data headers, the node index vector associated with each different classification data header is confirmed, and the storage location of the classification data is quickly locked based on the node index vector and the associated characteristic vector circle; If there is only a single item of classified data, the index output is performed directly; If there are multiple different classification data, the byte quantity associated with the different classification data is determined based on the attributes of the corresponding classification data and marked as Z k , where k represents different classification data and is based on the byte size Z k Sort multiple different classification data in a way of ascending from small to large, and index and output the sorted classification data in turn; Specifically, this method of sequential indexing and outputting can effectively ensure the efficiency of indexing output while ensuring the sequential nature of data indexing, preventing data confusion and ensuring the overall effect of the data indexing process.

[0013] Some of the data in the above formulas are numerically calculated by removing their dimensions. Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0014] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A financial investment and financing management platform based on artificial intelligence and big data, characterized by: include: The financial data acquisition end collects financial data generated by different data sources within a specified period, where the specified period is a preset period; The multi-threaded input control terminal inputs the financial data collected in this batch based on the preset multiple threads, inputs the financial data of the specified data type according to the inputtable type associated with the corresponding thread, and adjusts the computing power resources associated with a single thread based on the input characteristics generated in real time, so that multiple threads are in a load-balanced state when inputting synchronously; The cloud database stores the financial data entered in the current batch, where each different classification data in the financial data is associated with a different storage node path; The index vector construction end confirms the storage node path of each different classification data in the current batch, and generates a corresponding number of hierarchy circles according to the hierarchical classification of multiple groups of different storage node paths. Then, according to the total number of storage node paths, the hierarchy circles of the specified level are confirmed to have equal points. Then, the multiple groups of equal points associated with a single group of classification data are used to lock the node index vector associated with the corresponding classification data. The index output end receives the index instruction input from the outside, and confirms the associated node index vector from the storage unit based on the index instruction, and then directly determines the storage location of the corresponding classification data from the cloud database based on the associated storage node path, and directly performs index output.

2. The financial investment and financing management platform based on artificial intelligence big data according to claim 1 is characterized in that: The specific method of the multi-threaded input control end to adjust the computing power resources in real time is: Based on the financial data collected in this batch, classified data of different data types are identified, and each group of classified data corresponds to one data type; According to the preset multiple threads and the enterable types recorded by the corresponding threads, the enterable types are preset types, corresponding to the data types, matching the classification data consistent with the corresponding thread types, and recording the classification data; Record the input rate associated with each thread in real time and mark it as V i , where i represents different threads, and the preset standard input rate of each thread is calibrated as B i , where the standard input rate is the preset rate, and real-time confirmation is available: (B i -V i )>Y1, if any, then this thread is marked as the thread to be adjusted, where Y1 is the preset fluctuation value; Based on the calibrated threads to be adjusted and other actually running threads, confirm the computing resources associated with each thread, and allocate the computing resources associated with other threads to the threads to be adjusted. Each time the computing resources associated with the other thread allocation process are one unit, the unit is the preset unit, that is, 1FLOPS. Confirm the thread load balancing associated with each allocation process: Input rate V based on real-time monitoring i and standard input rate B i , confirm the input feature ratio associated with the corresponding thread: TB i =V i ÷B i , and compare the multiple sets of input features associated with multiple threads to TB i Perform variance processing, confirm the standard variance, and confirm whether the standard variance meets the following requirements: standard variance ≤ Y2, where Y2 is the preset value. If so, the allocation process of computing resources is terminated. If not, the allocation is continued and the threads to be adjusted are confirmed in real time until multiple threads are load balanced.

3. The financial investment and financing management platform based on artificial intelligence big data according to claim 2 is characterized in that: If it does not exist (B i -V i )>Y1, no calibration is performed.

4. The financial investment and financing management platform based on artificial intelligence big data according to claim 1 is characterized in that: In the cloud database, financial data entered in different batches are stored in different classification intervals.

5. The financial investment and financing management platform based on artificial intelligence big data according to claim 1 is characterized in that: The specific method of locking the node index vector associated with the corresponding classification data at the index vector construction end is: Based on the storage node path of each different classified data, the first group of hierarchical paths is used as the initial path, and the other node paths that appear successively are confirmed in turn. With the initial path as the center of the circle, a hierarchical circle associated with the second group of hierarchical paths is generated, and based on the multiple groups of storage node paths included in the second group of hierarchical paths, the total number G of different storage node paths is identified from the multiple groups of storage node paths, and then the number of evenly divided points of the hierarchical circle associated with the second group of hierarchical paths is confirmed, and the number of evenly divided points is also G, and the storage node path associated with each different evenly divided point is recorded; Generate the hierarchical circles associated with the third hierarchical path, the latter hierarchical circles are located outside the former hierarchical circles, and the same determination method as the hierarchical circles of the second hierarchical path is used to confirm the equal division points on the circumference of the third hierarchical circle; And so on, until the level circles associated with the last set of level paths and the internal equal division points are confirmed, the generated multiple sets of level circles are recorded as the characteristic vector circles associated with this batch of financial data; According to the storage node path of the classification data associated with the corresponding classification data, the associated equal-division points are selected on the corresponding level circle in turn, and line segments are connected in turn from the center of the feature vector circle outward to generate a node index vector belonging to this classification data.

6. The financial investment and financing management platform based on artificial intelligence big data according to claim 5 is characterized in that: The node index vectors of all the classified data in this batch are confirmed in the feature vector circle, and the confirmed classified data headers are matched one by one with the corresponding node index vectors to generate an index table, and the generated index table and the feature vector circle associated with the current batch are transmitted to the storage unit for storage.

7. The financial investment and financing management platform based on artificial intelligence big data according to claim 6 is characterized in that: The specific method of index output at the index output end is as follows: Based on the received index instruction, confirm the classification data header associated with the index instruction in the storage unit index table, and then confirm the node index vector associated with the classification data header; Based on the node index vector and the associated characteristic vector circle, the storage location of the classification data is quickly locked. If a single index instruction corresponds to multiple different classification data headers, the node index vector associated with each different classification data header is confirmed, and the storage location of the classification data is quickly locked based on the node index vector and the associated characteristic vector circle; If there is only a single item of classification data, the index output is performed directly.

8. The financial investment and financing management platform based on artificial intelligence big data according to claim 7 is characterized in that: If there are multiple different classification data, the byte quantity associated with the different classification data is determined based on the attributes of the corresponding classification data and marked as Z k , where k represents different classification data and is based on the byte size Z k Sort multiple different classification data from small to large, and index and output the sorted classification data in turn.

Images