TWI829372B - Funding demand forecasting system and forecasting method - Google Patents

Abstract

A method for predicting capital demand, comprising capturing the pending iris feature of a user with an iris recognition device, and determining whether there is a matching registered iris feature. If it exists, the user's identity is recorded as an authenticated identity; if not, the user's identity is recorded as a suspicious identity. When the identity is an authenticated identity, the computing device is notified by means of a message, so as to present the preset data on the display, and the preset data correspond to the data types, respectively. Afterwards, the computing device transmits a signal to notify the eye tracking device to track the user's line of sight, so as to lock the target data in the preset data. The locked target data is transmitted to the computing device, and the computing device obtains the target weight according to the corresponding data type. The computing device performs a weighted sum of the target data and the target weights to obtain an evaluation score and predict the funding requirement.

Description

Fund demand forecasting system and forecasting method

The invention relates to a capital demand prediction system and a prediction method, in particular to a supplier capital demand prediction system and a prediction method.

For banks, in addition to long-term trend predictions for specific indicators, capital flows also require dynamic analysis of suppliers' short-term capital needs. In particular, suppliers' funding needs will fluctuate with the economic environment or various factors, but banks can often only passively wait for suppliers to request funding before providing assistance, and are unable to provide corresponding services efficiently.

In view of the above, the present invention provides a supplier fund demand forecasting system and forecasting method.

A fund demand forecasting method according to an embodiment of the present invention includes using an iris recognition device to capture a set of undetermined iris features of a user, and determining whether there is a set of registered iris features consistent with the set of undetermined iris features. If there is a set of logged-in iris features consistent with the set of pending iris features, the iris recognition device records the user's identity as an authentication identity corresponding to the logged-in iris features; if there is no set of logged-in iris features consistent with the set of pending iris features, If the iris feature is logged in, the user's identity is recorded as a suspicious identity corresponding to the undetermined iris feature. point. When the iris recognition device records the user's identity as the authenticated identity, it notifies a computing device, and the computing device presents a plurality of preset data on a display, and the preset data respectively corresponds to a plurality of data categories. After the display displays the default data, the computing device signals an eye-tracking device to track the user's gaze to lock multiple target data in the default data. The eye tracking device transmits the locked target data to the computing device, and the computing device obtains respective target weights based on the data categories corresponding to the target data. The computing device performs a weighted sum of the target data and the corresponding target weights to obtain an evaluation score; and calculates a predicted fund requirement based on the evaluation score.

With the above structure, the fund demand forecasting system and forecasting method disclosed in this case can perform biometric verification on a user through an iris recognition device to ensure that the user's identity is correct, and determine the target data on which the user's eyes are looking through an eye-tracking device. To select the funding information related to the supplier, and then perform weight calculation on the selected information through the computing device, so the forecast funding requirements of the supplier can be obtained based on multiple pieces of data from the supplier. In this way, users or banks can prepare an appropriate amount of funds in advance and provide corresponding assistance and services efficiently.

The above description of the present disclosure and the following description of the embodiments are used to demonstrate and explain the spirit and principles of the present invention, and to provide further explanation of the patent application scope of the present invention.

1,1’: Funding demand forecasting system

10: Iris recognition device

20:Computing device

30:Display

40: Eye tracking device

41: Infrared emitter

42:Infrared camera

50:Finger vein identification device

D: sight direction

E: Eyeball

I:iris

P:pupil

S: sclera

S100~S108, S201~S209, S301~S303: steps

FIG. 1 is a block diagram of a fund demand forecasting system according to an embodiment of the present invention.

FIG. 2 is a schematic diagram of iris recognition of a fund demand prediction method according to an embodiment of the present invention.

FIG. 3 is a schematic diagram of an eye tracking device of a fund demand prediction system according to an embodiment of the present invention.

FIG. 4 is a block flow chart of a method for forecasting capital needs according to an embodiment of the present invention.

FIG. 5 is a schematic diagram of selecting target data of the fund demand forecasting system according to an embodiment of the present invention.

FIG. 6 is a block diagram of a fund demand forecasting system according to another embodiment of the present invention.

FIG. 7 is a block flow chart of a method for forecasting capital needs according to another embodiment of the present invention.

FIG. 8 is a block flow chart of a method for forecasting capital requirements according to another embodiment of the present invention.

The detailed features and advantages of the present invention are described in detail below in the implementation mode. The content is sufficient to enable anyone skilled in the relevant art to understand the technical content of the present invention and implement it according to the content disclosed in this specification, the patent scope and the drawings. , anyone familiar with the relevant art can easily understand the relevant objectives and advantages of the present invention. The following examples further illustrate the aspects of the present invention in detail, but do not limit the scope of the present invention in any way.

Please refer to FIG. 1 , which is a block diagram of a fund demand forecasting system according to an embodiment of the present invention. As shown in Figure 1, the capital demand forecasting system 1 includes an iris recognition device There is a device 10, a computing device 20, a display 30 and an eye tracking device 40, wherein the computing device 20 serves as a signal processing center and is connected to other devices. It should be understood that although the computing device 20 in this example is connected to other devices, and the other devices are each connected through the computing device 20 , in other embodiments, between other devices (such as the display 30 and the eye tracking device 40 ) may have a direct signal connection relationship, and this case is not limited to this.

Please refer to FIG. 2 in conjunction with FIG. 1. FIG. 2 is a schematic diagram of iris recognition of a fund demand prediction method according to an embodiment of the present invention. In this example, the iris recognition device 10 can be equipped with a processor with machine vision capabilities (such as a central processing unit, a microcontroller or a programmable logic controller) with a dedicated camera for capturing a set of images of a user. Pending iris features, and determine whether there is a set of login iris features consistent with the set of pending iris features. For example, the iris features may include iris texture, iris color, the boundary between iris I and pupil P, and the boundary between iris I and sclera S. Specifically, approximately 240 feature points can be found on the iris of an eye, that is, each person's iris has a unique pattern for identification by the iris recognition device 10 .

Please refer to FIG. 3 in conjunction with FIG. 1 . FIG. 3 is a schematic diagram of an eye tracking device of a fund demand prediction system according to an embodiment of the present invention. In this example, the eye tracking device 40 is used to track the user's line of sight to lock the target data in the data selection process. For example, the eye tracking device 40 may be a wearable device (such as glasses) having a processor (such as a central processing unit, a microcontroller or a programmable logic controller) to better track the movement of the user's eyeballs. As shown in FIG. 3 , the eye tracking device 40 may include an infrared emitter 41 and an infrared camera 42 , which may be used to irradiate infrared light to the surface of the eyeball and receive infrared light reflected from the surface of the eyeball, respectively. Specifically, the principle of eye tracking can be based on the iris I, as shown in Figure 2. The reflectivity of the pupil P and the sclera S to infrared light is different so that the position of the pupil P or the iris I can be identified to calculate the user's line of sight direction. The relevant details will not be described here if those with ordinary knowledge in the art can understand them. In addition, in other embodiments, the eye tracking device 40 may not be a wearable device but an external device, such as a processor with machine vision capabilities and a dedicated camera to identify the user's gaze direction. This case is not limited thereto.

Please refer to FIG. 4 in conjunction with FIG. 1 . FIG. 4 is a block flow chart of a method for forecasting capital needs according to an embodiment of the present invention. As shown in Figure 4, the method for predicting financial requirements includes step S100: using the iris recognition device 10 to capture a set of undetermined iris features of the user; step S101: using the iris recognition device 10 to determine whether there is a set of undetermined iris features that match the user's set of undetermined iris features. Register the iris feature. If it does not exist, go to step S102. If it exists, go to step S103. Step S102: The iris recognition device 10 records the user's identity as a suspicious identity corresponding to the undetermined iris feature. Step S103: The iris recognition device 10 The user's identity is recorded as an authentication identity corresponding to the logged-in iris feature. Step S104: The display 30 presents a graphical user interface and shows a plurality of preset data. Step S105: The eye tracking device 40 tracks the user's line of sight to Lock the target data in the default data. Step S106: The computing device 20 obtains multiple target weights according to the data type of the target data. Step S107: The computing device 20 performs a weighted sum of the target data and the corresponding target weight to obtain an evaluation score. Step S108: The computing device 20 calculates a predicted fund requirement based on the evaluation score.

Examples and detailed descriptions of each step shown in Figure 4 are given below. In step S100, the iris recognition device 10 can capture an image of the user's eyes, which may include pupils, iris, sclera, etc., and then capture the undetermined iris features of the eye image through the internal processing unit. In step S101, the iris recognition device 10 may use the undetermined iris characteristics obtained in step S100 to Compare with multiple sets of internally pre-stored login iris features to determine whether there is a set of login iris features consistent with the set of pending iris features. In one implementation, the compliance rate of definable conditions and features needs to reach 90%, but this case is not limited to this. It should be noted that if there are multiple sets of registered iris features that match the set of pending iris features, the judgment of "there is a set of iris features consistent with the pending iris features" is still not met, that is, it is judged that there is only one set that matches. Login iris features.

In step S102, since it cannot properly determine which registered iris feature in the database the user's pending iris feature corresponds to, the iris recognition device 10 records the user's identity as a suspicious identity in the built-in memory of the computing device 20 in the body. In step S103, the iris recognition device 10 records the user's identity as the authentication identity in the built-in memory of the computing device 20, and notifies the computing device 20. The computing device 20 then transmits a plurality of default data to the display 30 . In step S104 , the display 30 receives the default data from the computing device 20 and displays it on a graphical user interface and returns a signal to the computing device 20 . In step S105 , the computing device 20 notifies the eye-tracking device 40 to track the user's line of sight, and the eye-tracking device 40 returns the user's line of sight to the computing device 20 to lock the graphic where the extension line of the line of sight intersects on the display 30 The target data in the default data on the user interface. In step S106, the computing device 20 obtains a plurality of target weights according to a plurality of data categories respectively corresponding to the target data. In step S107, the computing device 20 performs a weighted sum of the target data and the corresponding target weights to obtain an evaluation score. In step S108, the computing device 20 calculates a predicted fund requirement based on the evaluation score. For example, if a supplier is evaluated in each month and it is found that its evaluation score in May is the highest, a reserved fund can be appropriately allocated in May according to a proportion as the supplier's capital demand forecast.

In an implementation form, the above target data may include an upstream manufacturer's purchase transaction data, corresponding to the first weight, a central factory transaction data, a second weight, a credit information, a third weight, a financial data, corresponding to the fourth weight, and an external economic data, corresponding to the fifth weight. An example is given below to illustrate.

The first weight corresponding to the purchase and transaction situation of the upstream manufacturer is set to 30 points (total score is 100 points), of which "appointment information upload/number of orders per month" accounts for 9 points, and "appointment transfer/voucher amount" accounts for 12 points. "Transfer/Number of delivery dates" accounts for 9 points. The calculation method is as follows Table 1.

Set the second weight corresponding to the central factory's transactions to 25 points, of which "appointment information upload/number of orders per month" accounts for 7.5 points, "voucher amount" accounts for 10 points, and "transfer/number of delivery days" accounts for 7.5 point. The calculation method is as follows in Table 2.

Set the third weight corresponding to the credit situation to 20 points, of which "approved amount of quota" accounts for 2 points, "amount of automatic transfers for the month" accounts for 10 points, "number of transfers for the month" accounts for 4 points, "average transfer interest rate and "Comparison with peers" accounted for 4 points. The calculation method is as follows in Table 3.

Set the fourth weight corresponding to the customer's financial status to 15 points, of which "net worth ratio" accounts for 3 points, "revenue growth rate" accounts for 3 points, "quick ratio" accounts for 3 points, and "cash flow ratio" accounts for 3 points "Increase or decrease in loan balance of Lianzheng Center" accounts for 3 points. The calculation method is as follows in Table 4.

The fifth weight corresponding to the economic and industrial situation is set to 10 points, of which "industrial weather distribution overview" accounts for 2 points, "purchasing managers index" accounts for 2 points, "producer price index" accounts for 2 points, and "industry/finance" News" accounted for 4 points. The calculation method is as shown in Table 5 below.

After predicting the scores of each factor listed in the above table, the scores for the current month can be calculated as Table 6.

Furthermore, the situation in the past five years can be analyzed, as shown in Table 7 below.

Therefore, the assessment score of the supplier's capital needs in each month can be calculated to predict the monthly location of future capital needs, allowing banks to respond as early as possible, as shown in Table 8 below.

It can be understood from the above examples that there are many factors that affect a supplier's capital requirements. Therefore, in another embodiment, a machine learning algorithm can be used to adjust the weight of each target to optimize the prediction accuracy of capital requirements. The steps of determining the target weight by the computing device include: obtaining a plurality of historical fund requirements and obtaining a plurality of historical data. Each of the historical data includes a historical upstream manufacturer purchase transaction sub-data and a historical central factory transaction transaction sub-data. , a historical credit sub-data, a historical financial sub-data and a historical external economic sub-data, the computing device performs the weighted sum of the historical data and a plurality of preset modulation weights to obtain a test fund requirement, and obtain the test A difference between the capital demand and the historical capital demand, wherein the preset adjustment weights respectively correspond to the historical upstream manufacturer purchase transaction sub-data, the historical central factory transaction transaction sub-data, a historical credit sub-data, and a historical financial sub-data Data and a historical external economic sub-data, the computing device adjusts the preset modulation weights through a gradient descent algorithm, and passes the historical data and the adjusted preset modulation weights through the above-mentioned The weighted summation recalculates another difference until the other difference reaches a minimum value to obtain a plurality of confirmed modulation weights, and records the confirmed modulation weights as the target weights.

For example, historical funding requirements and historical data include January data for the past five years. Initially, the error value of the capital demand can be calculated using the preset adjustment weights and historical data, and then the gradient descent algorithm is used to adjust each weight to reduce the error value until the error value reaches a minimum value, and these adjustment weights are recorded. is the target weight. That is to say, for January in the past five years, the calculated target weights can be a set of target weights that can better meet the funding demand situation in the past five years. In addition, in addition to the gradient descent method, other optimization algorithms can also be used to obtain different convergence speeds or accuracy rates, such as stochastic gradient Stochastic gradient descent (SGD), adaptive learning rate method (Adaptive Learning Rate Method) or conjugate gradient method (Conjugate Gradient Method), etc. This case is not limited to these.

On the operation interface, please refer to FIG. 5 , which is a schematic diagram of selecting target data of the fund demand forecasting system according to an embodiment of the present invention. As shown in FIG. 5 , after the user completes the identity recognition, the user can start selecting target data through the eye tracking device 40 and the display 30 , that is, the eye tracking device 40 tracks the sight direction D of the user's eye E to intersect on the graphical user interface. The target data corresponding to the point completes the data selection operation.

Please refer to FIG. 6 , which is a block diagram of a fund demand forecasting system according to another embodiment of the present invention. In this example, the fund demand prediction system 1', in addition to the above-mentioned iris recognition device 10, computing device 20, display 30 and eye tracking device 40, also includes a set of undetermined finger vein characteristics for acquiring the user and determining the undetermined Whether the finger vein characteristics match the registered finger vein characteristics corresponding to the user's identity, and the finger vein used to measure the blood flow speed of a finger vein of the user and determine whether the blood flow speed of the finger vein is greater than a preset speed The identification device 50 is signal-connected to the computing device 20 . The finger vein recognition device 50 has a processor (such as a central processing unit, a microcontroller or a programmable logic controller) to perform the above determination. In addition, the capital demand forecasting system 1' in this example may also include a special contact lens (not shown) for the user to wear, using an artificial totem of the special contact lens itself plus an iris of the user The pattern generates registered iris features pre-stored in the iris recognition device 10 . That is to say, the set of undetermined iris features corresponds to the iris image of the user after wearing a special contact lens, and the set of registered iris features pre-stored in the iris recognition device includes an iris pattern of the user plus an iris pattern. An artificial totem of special contact lenses, in which the iris recognition device 10 is further used A superimposed pattern of the artificial totem plus the user's iris pattern is obtained in advance during a registration stage as the set of registered iris features. This prevents malicious users from forging their identity by wearing contact lenses with special patterns.

Please refer to FIG. 7 in conjunction with FIG. 4 and FIG. 6 . FIG. 7 is a block flow chart of a fund demand forecasting method according to another embodiment of the present invention. As shown in FIG. 7 , after step S104 (the display 30 displays the preset data), the computing device 20 may determine whether the eye tracking device 40 cannot lock the target data of eye gaze for a preset time such as 20 seconds (step S201 ). If not, proceed to step S105, where the eye-tracking device 40 tracks the user's line of sight to lock the target data in the default data; if yes, the computing device 20 may notify the display to open an interactive teaching page (step S202). Specifically, the inability of the user's eyes to lock the target data may mean that the user does not understand the operation method or the user is looking at other devices at the same time.

Then, the computing device 20 notifies the finger vein recognition device 50 to intercept a set of undetermined finger vein features of the user and determine whether the set of undetermined finger vein features match a registered finger vein feature corresponding to the identity of the user ( Steps S203 and S205), and measure the blood flow speed of a finger vein of the user and determine whether the blood flow speed of the finger vein is greater than a preset speed (steps S203 and S204). Based on the above two judgments, further check whether the user's identity is suspicious, that is, if the registered finger vein characteristics do not match the to-be-determined finger vein characteristics, or if the finger vein blood flow speed is greater than the preset speed, the finger vein identification device 50 records The user's identity is a suspicious identity (step S206). Once the user's identity is recorded as a suspicious identity, a usage authority of the user is stopped and all information provided by the user is recorded (step S207).

On the other hand, if the finger vein recognition device 50 determines that the registered finger vein characteristics match the undetermined finger vein characteristics, the user's identity is recorded as a two-factor authentication identity (step S208). Therefore, as long as the user's identity is not recorded as a suspicious identity, the computing device 20 will allow the user to continue the operation (step S209). It should be noted that the "suspicious identity" and "authentication identity" in this example can exist in the user's identity at the same time. However, the computing device 20 determines whether to suspend the user's use authority based on "whether the user has a suspicious identity." . In other words, a user with normal finger vein characteristics may have his or her authority suspended due to abnormal finger vein blood flow speed, thereby preventing users with reasonable identities from performing abnormal operations.

Please refer to FIG. 8 in conjunction with FIG. 4. FIG. 8 is a block flow chart of a method for forecasting capital needs according to another embodiment of the present invention. As shown in FIG. 8 , after the iris recognition device 10 captures the user's undetermined iris features (ie, step S100 ), the iris recognition device 10 can determine whether there is no dynamic change in the undetermined iris features within a preset period (step S301 ). If yes, the iris recognition device 10 records the user's identity as a suspicious identity, stops the user's use authority and records all the information provided by the user (step S303); if not, the computing device 20 allows the user to continue. operation (step S101), and then enter the above-mentioned step S101.

With the above structure, the supplier's funding demand forecasting system and forecasting method disclosed in this case can perform multiple verifications on a user through the iris recognition device and the finger vein recognition device to ensure that the user's identity and operation are correct, and through the eye tracking device Determine the target data that the user is looking at to select financial information related to the supplier, and then perform weighting calculations on the selected information through the computing device, so that the information can be obtained based on multiple pieces of data from the supplier. gold forecast demand. In this way, users or banks can prepare an appropriate amount of funds in advance and provide corresponding assistance and services efficiently.

Although the present invention is disclosed in the foregoing embodiments, they are not intended to limit the present invention. All changes and modifications made without departing from the spirit and scope of the present invention shall fall within the scope of patent protection of the present invention. Regarding the protection scope defined by the present invention, please refer to the attached patent application scope.

S100~S108: steps

Claims (8)

A fund demand forecasting method includes: using an iris recognition device to capture a set of undetermined iris features of a user; using the iris recognition device to determine whether there is a set of login iris features consistent with the set of undetermined iris features; if there is a set of login iris features consistent with the set of undetermined iris features; If there is a set of logged-in iris features that match the set of pending iris features, the iris recognition device records an identity of the user as an authentication identity corresponding to the logged-in iris features; if there is no set of logged-in iris features that match the set of pending iris features, Feature, the iris recognition device records the identity of the user as a suspicious identity corresponding to the undetermined iris feature; when the iris recognition device records the identity of the user as the authenticated identity, a notification is sent to a computing device that the The computing device presents a plurality of preset data on a display, and the preset data respectively corresponds to a plurality of data categories; after the display displays the preset data, the computing device signals an eye tracking device to track the The user's line of sight, the eye tracking device returns the user's line of sight to the computing device to lock a plurality of target data in the preset data; the computing device respectively corresponds to a plurality of data categories according to the target data , obtain multiple target weights; the computing device performs a weighted sum of the target data and the corresponding target weights to obtain an evaluation score; and the computing device calculates a predicted fund requirement based on the evaluation score. The fund demand forecasting method as described in claim 1, wherein the target weights include: - a first weight, corresponding to an upstream manufacturer's purchase and transaction data; - a second weight, corresponding to a central factory's transaction data; - a third weight, corresponding to a credit information; - a fourth weight, corresponding to a financial data; and a fifth weight corresponding to an external economic data. The fund demand forecasting method as described in claim 1, wherein the target weights are determined by the computing device executing a plurality of steps, and the steps include: obtaining a plurality of historical fund demands; obtaining a plurality of historical data. Each of the data includes a historical upstream manufacturer purchase transaction sub-data, a historical central factory transaction transaction sub-data, a historical credit sub-data, a historical financial sub-data and a historical external economic sub-data; combine these historical data with The weighted summation of multiple preset adjustment weights is performed to obtain a test capital requirement, and a difference between the test capital requirement and the historical capital requirement is obtained, wherein the preset adjustment weights respectively correspond to the historical upstream manufacturer purchase transactions. sub-data, the historical central factory transaction sub-data, a historical credit sub-data, a historical financial sub-data and a historical external economic sub-data; adjust the default adjustment weights through a gradient descent algorithm, and combine these The historical data and the adjusted preset modulation weights are recalculated through the above-mentioned weighted sum until the other difference reaches a minimum value to obtain multiple confirmed modulation weights; and the confirmed modulation weights are recorded as The weight of these goals. The fund demand prediction method as described in claim 1, wherein the set of undetermined iris features includes an iris pattern, an iris color, and a boundary between the iris and the pupil and sclera. The fund demand forecasting method as described in claim 1 further includes: if the eye tracking device cannot lock the target data for which the eye is fixed for a preset time, the computing device notifies the display to open an interactive teaching page. The fund demand forecasting method as described in claim 5, wherein when the interactive teaching page is opened, the computing device notifies the user to perform verification through the interactive teaching page, and the computing device notifies a finger vein recognition device to execute: capture Get a set of undetermined finger vein features of the user and determine whether the set of undetermined finger vein features match a registered finger vein feature corresponding to the identity of the user; if the set of registered finger vein features matches the set of undetermined finger vein features If the characteristics match, record the identity of the user as a two-factor authentication identity; if the set of logged-in finger vein characteristics does not match the set of undetermined finger vein characteristics, record the identity of the user as the suspicious identity; measure the user The blood flow speed of a finger vein and determine whether the blood flow speed of the finger vein is greater than a preset speed; if the blood flow speed of the finger vein is greater than the preset speed, record the identity of the user as a suspicious identity; and if If the identity of the user is not recorded as the suspicious identity, the computing device allows the user to continue operating, otherwise, the user's usage authority is terminated and all information provided by the user is recorded. The fund demand forecasting method as described in claim 1, wherein the set of undetermined iris features corresponds to the user's iris image after wearing a special contact lens, and the iris recognition The set of registered iris characteristics pre-stored in the recognition device includes an iris pattern of the user plus an artificial totem of a special contact lens. The fund demand prediction method as described in claim 1 further includes: if the iris recognition device determines that the set of pending iris features have no dynamic changes within a preset period of time, recording the identity of the user as a suspicious identity; And if the identity of the user is not recorded as the suspicious identity, the computing device allows the user to continue operating, otherwise, the user's usage authority is terminated and all information provided by the user is recorded.

Images