TWI760254B - Automatic fund depositing method using neural network - Google Patents

Abstract

An automatic fund depositing method using a neural network comprises collecting, by an account monitoring host, a current deposit amount of an account every predetermined period, driving, by a transfer management host, a machine learning model to calculate an amount difference between the current deposit amount and a fixed amount and compare the amount difference with a lower limit of the amount difference and an upper limit of the amount difference for generating a deposit transfer amount during every predetermined period.

Description

Automated Savings Method Using Neural-like Networks

The present invention relates to a financial transaction management method, especially a time deposit management method.

In the existing automatic time deposit system, customers set a fixed amount of time deposit according to their actual income for the time period of time deposit, so as to prepare for the future education fund, travel fund or retirement fund. When the fixed deposit amount set is too low, it will be difficult for wealth management personnel to persuade customers to convert the fixed deposit to purchase other wealth management products, and the expense of fixed deposit interest will cause the burden of bank costs. Conversely, when the set time deposit amount is too high, if the customer temporarily needs funds and must cancel the contract, although the bank will obtain the additional cancellation fee paid by the customer, it will also lose the entire fixed deposit amount, resulting in financial personnel unable to Customers provide appropriate financial products to help customers earn higher returns than savings interest.

In view of this, there is a real need for an improved automatic savings system that can at least address the above deficiencies.

The present invention is to provide an automatic saving method using a neural network, which can adjust the customer's current fixed deposit and transfer amount according to the customer's actual income and expenditure status.

According to an embodiment of the present invention, an automatic deposit method using a neural network-like network is disclosed, comprising: collecting an existing deposit data of an account in each predetermined cycle by an account monitoring host, the existing deposit data including a deposit amount and a deposit amount. Deposit time; execute a machine learning model with the transfer management host, the machine learning model is a type of neural network, and receive a fixed amount, an upper limit of the amount difference, a lower limit of the amount difference and the deposit amount by the type of neural network; The neural network-like calculates the difference between the deposit amount and a fixed amount; the neural network compares the difference with a lower limit and an upper limit of the difference respectively; if the neural network determines that the difference is greater than zero and less than or equal to the lower limit of the amount difference, output a current fixed deposit and transfer amount by this type of neural network, and the current fixed deposit and transfer amount is less than the fixed amount; if the type of neural network determines that the amount difference is greater than the amount Difference upper limit, output the current fixed deposit and transfer amount by this type of neural network, and the current fixed deposit and transfer amount is greater than the fixed amount; When the difference between the deposit and transfer amount is less than 0, the neural network is used to increase the first weight, increase the second weight and decrease the third weight according to the fixed deposit and transfer amount difference and through a backward pass algorithm. weight, wherein a first adjustment percentage of the first weight is smaller than a second adjustment percentage of the second weight and a third adjustment percentage of the third weight; if the current fixed deposit and transfer amount and the historical fixed deposit and transfer amount When the difference between the fixed deposit and transfer amount falls within a preset amount range, the neural network maintains the first weight, the second weight and the third weight; if the fixed deposit and transfer amount difference is greater than 0 , then the neural network is used to increase the first weight, increase the second weight and increase the third weight according to the difference between the fixed deposit and transfer amount and through the backward pass algorithm, wherein the first adjustment The percentage is greater than the second adjustment percentage and the third adjustment percentage; if the neural network determines that the difference in the amount is less than 0, the neural network does not transfer funds; a drawing circuit of the transfer management host is used to manage the transfer according to a A current fixed deposit and transfer amount of a year generates a fixed deposit amount curve of the current year, and transmits the fixed deposit amount curve to a customer's e-mail through a network; and receives a password to reset the fixed amount, the amount The magnitude of the difference between the upper limit of the difference and the lower limit of this amount.

The automatic saving method using a neural network disclosed in the present invention can plan a customized fixed deposit and transfer amount and perform automatic fixed deposit and transfer according to the actual personal financial income and expenditure status of the customer. When the customer has a high demand for funds, the automatic savings system automatically reduces the customer's fixed deposit and transfer amount. On the contrary, when the customer's funds are relatively abundant, the automatic savings system automatically increases the customer's fixed deposit amount, so the customer does not need to adjust it. Fixed deposit transfer amount. To achieve the effect of simplifying the operation process.

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

The detailed features and advantages of the present invention are described in detail in the following embodiments, and the content is sufficient to enable any person skilled in the relevant art to understand the technical content of the present invention and implement it accordingly, and according to the content disclosed in this specification, the scope of the patent application and the drawings , any person skilled in the related art can easily understand the related objects and advantages of the present invention. The following examples further illustrate the viewpoints of the present invention in detail, but do not limit the scope of the present invention in any viewpoint.

FIG. 1 is a functional block diagram of an automatic saving system according to a first embodiment of the present invention. As shown in FIG. 1 , the automatic savings system 100 includes an account monitoring host 10 and a transfer management host 12 . The account monitoring host 10 communicates with the internal server S of the bank through a wired or wireless network. The internal server S of the bank stores the financial transaction data of all customers of the bank. The financial transaction data includes deposit data (at least the deposit amount and date of deposit), withdrawal data (at least the withdrawal amount and date of withdrawal), and transfer data. (At least include the transfer amount, transfer date and transfer account). The account monitoring host 10 is used for collecting deposit data and transfer data of the customer's account every predetermined period, such as one month. The transfer management host 12 includes a processor 121 and an external memory 122 , and the processor 121 is electrically connected to the external memory 122 . The external memory 122 is used to store the deposit data about the customer collected by the account monitoring host 10, the preset fixed amount, the difference between the deposit amount and the fixed amount, the preset upper limit and lower limit of the difference, and financial transfer data, etc. , where the fixed amount is the fixed deposit and transfer amount of each month preset by the customer in advance. The processor 121 communicates with the account monitoring host 10 through a wired or wireless network. The processor 121 further has an internal memory 123 , and the internal memory 123 stores a machine learning model 13 . The machine learning model 13 includes a neural network, a fuzzy logic model, a hidden Markov model, a decision tree, a Bayesian algorithm, a conditional random field or a support vector machine. In one embodiment, the machine learning model 13 is, for example, a class Neural network, and quasi-neural network can use the historical deposit data, historical transfer information, preset fixed amount, upper limit and lower limit of difference difference collected by the account monitoring host 10 for offline training in advance, so that the quasi-neural network can directly When going online, when receiving the real-time existing deposit information, it can calculate the fixed deposit transfer amount that best meets the customized needs. The transfer management host 12 is used to execute the machine learning model 13 to drive the trained machine learning model 13 to calculate the difference between the customer's deposit amount and the fixed amount, and compare the amount difference with the lower limit of the difference and the upper limit of the difference, according to As a result of the comparison, a certain deposit and transfer amount is generated every predetermined period. In this embodiment, the predetermined period is every month, so that the machine learning model 13 of the automatic savings system 100 adjusts the current fixed deposit and transfer amount every month according to the actual income and expenditure status of the customer. The transfer management host 12 further includes a drawing circuit 124. After the machine learning model 13 generates multiple pieces of data about the current fixed deposit and transfer amount, the drawing circuit 124 can generate a certain deposit amount according to the multiple current fixed deposit and transfer amounts and their corresponding transfer times. and the fixed deposit amount curve is stored in the external memory 122 . For example, the machine learning model 13 generates the current fixed deposit and transfer amount for a year, then the drawing circuit 124 will generate a current year's fixed deposit amount curve according to the current fixed deposit and transfer amount of the whole year, thereby showing the customer's current fixed deposit amount curve. The fixed deposit transfer amount for each month of the year.

Furthermore, the transfer management host 12 can communicate with the client's mobile device or personal computer through a wireless network, so as to transmit the data of the fixed deposit amount curve to the client's e-mail. In addition, for the consideration of information security, it can also be set that the client's mobile device or personal computer needs to additionally install a designated application program matched with the transfer management host 12 to communicate with the transfer management host 12 to receive the fixed deposit amount curve data. .

FIG. 2 is a functional block diagram of an automatic saving system according to a second embodiment of the present invention. As shown in FIG. 2, the difference between the automatic saving system 200 of the second embodiment and the automatic saving system 100 of the first embodiment is that the automatic saving system 200 further includes an interest rate management host 20, and the interest rate management host 20 is connected through a wired or wireless network. Communication and connection with the processor 121 of the transfer management host 12 and the account monitoring host 10, the interest rate management host 20 is used to determine whether the average value of the current fixed deposit and transfer amount of the customer in a predetermined period is greater than the fixed amount, thereby judging whether to adjust the customer's balance. Fixed deposit interest rate. For example, if the customer's current fixed deposit and transfer amount is higher than the fixed amount for more than six months in the current year, so that the average current fixed deposit and transfer amount of the current year is greater than the fixed amount, the interest rate of the automatic savings system 200 The management host 20 automatically increases the customer's fixed deposit interest rate in the next year, thereby achieving the effect of customizing the fixed deposit transfer amount.

The account monitoring host 10 is used for judging whether the existing deposit data of the account includes the deposit statement image files belonging to other banks in addition to the deposit data from the internal bank server S of the bank. If the judgment result is negative, follow-up processing will be carried out directly based on the existing deposit information of the bank. If the judgment result is affirmative, the predetermined position in the deposit statement image file is further identified and the deposit balance is obtained. Next, update the existing deposit information with the sum of the deposit balance and the bank's deposit information.

In one embodiment, the default position of the deposit amount in the deposit statement image file is obtained according to the following identification algorithm, and the identification algorithm includes the following steps. Step 1: Each deposit statement image file is converted to grayscale image by RGB color space image. Step 2: Convert the grayscale image into a binarized image using the adaptive threshold value. Step 3: Perform horizontal projection and vertical projection on the binarized image. Step 4: Remove the blank area around the binarized image to obtain an image file for identification. Step 5: Use the classification algorithm to use the decimal point as an indicator to determine the position of each decimal point in the identification image file. Step 6: Based on the indicator, obtain a digital string along the horizontal direction, wherein the digital string contains continuous Arabic numerals and commas or consecutive Arabic numerals. Step 7: Identify the keywords in the identification image file by optical character recognition (OCR), for example, the keywords are balance, balance, summary or total. Step 8: Determine the number of horizontal reference lines or vertical reference lines that pass through each number string, and use the number string with the largest number as the deposit amount in the deposit statement file, where the horizontal parameter line or vertical reference line passes through The location of the keyword.

In addition to the above identification algorithms, Fast Convolutional Neural Networks (FCNN), Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN) or Long Short-Term Memory Neural Networks (LSTM) are based on multiple deposit pairs The billing samples are obtained for training. Since the format of the deposit statement image file may change, and LSTM has the advantage of quickly learning new samples, in this embodiment, the LSTM is used to obtain the default deposit amount in the deposit statement image file. Location.

FIG. 3 is a flowchart of an automatic saving method according to the first embodiment of the present invention. As shown in FIG. 3 , in step 301 , the account monitoring host 10 collects existing deposit data about the customer in each predetermined period, and transmits the existing deposit data to the external memory of the transfer management host 12 via a wired or wireless network 122 , wherein the existing deposit information at least includes the deposit amount and the deposit date. In this embodiment, the predetermined period is set to be one month.

In step 302, the processor 121 of the transfer management host 12 executes the machine learning model 13 to drive the machine learning model 13 to calculate the difference between the deposit amount and the fixed amount according to the deposit amount and the fixed amount input into the machine learning model 13 , and store the difference data in the external memory 122 .

In step 303, the machine learning model 13 compares the amount difference with the lower limit and upper limit of the amount difference input into the machine learning model 13, respectively, and then selectively executes steps 304, 305, 306 or 307 according to the comparison results. In addition, the customer can also communicate with the transfer management host 12 through a designated application installed on the personal computer or mobile device, so as to reset the fixed amount input into the machine learning model 13 through the function page provided by Internet banking or mobile banking, The magnitude of the upper and lower amount difference. In addition, the transfer management host 12 can also be provided with a security mechanism. Even after connecting to the transfer management host 12 through a designated application, a password must be entered to reset the fixed amount, the upper limit of the difference and the lower limit of the difference to avoid non-bank customers. The data in the transfer management host 12 can be modified freely.

In step 304, if the machine learning model 13 determines that the amount difference is greater than zero and not greater than the lower limit of the amount difference, the machine learning model 13 generates a reduced deposit amount less than the fixed amount, and automatically reduces the deposit amount from the customer's current deposit account Transfer funds to the customer's time deposit account, and store the reduced deposit amount and the transfer date in the external memory 122, wherein the reduced deposit amount is the current fixed deposit transfer amount that matches the current customer's financial income and expenditure status. For example, the customer initially sets the fixed deposit and transfer amount for each month as 10,000 yuan, the minimum amount difference is 2,000 yuan, and the upper limit for the amount difference is 6,000 yuan, which means the fixed amount and the lower limit of the amount difference input into the machine learning model 13 The upper limit of the difference with the amount is 10,000 yuan, 2,000 yuan and 6,000 yuan respectively. When the deposit amount in the customer's current deposit account is 11,000 yuan, the deposit amount input into the machine learning model 13 is 11,000 yuan, and the machine learning model 13 is based on the input deposit amount, the fixed amount, the lower limit of the difference and the upper limit of the difference. Automatically reduce the fixed deposit transfer amount from 10,000 yuan to 9,000 yuan, and the reduced fixed deposit amount is 9,000 yuan. Therefore, after the customer's demand deposit account is automatically transferred by the machine learning model 13, the balance in the demand deposit account is still RMB 2,000, which will not be lower than the predetermined lower limit of the difference.

In step 305, if the machine learning model 13 determines that the amount difference is greater than the lower limit of the amount difference and not greater than the upper limit of the amount difference, the machine learning model 13 transfers a fixed amount from the customer's current deposit account to the customer's time deposit, and transfers the fixed amount of the transfer. and its transfer date are stored in the external memory 122 . For example, the customer initially sets the fixed deposit and transfer amount for each month to be 10,000 yuan, the minimum amount difference to be 2,000 yuan, and the upper limit to the amount difference to be 6,000 yuan. When there is 15,000 yuan left in the customer's current deposit account, the machine learning model 13 calculates that the difference is 5,000 yuan and determines that the difference is between the preset upper limit and lower limit of the difference, then the machine learning model 13 Automatically transfer a fixed amount (RMB 10,000) from the client's checking account to the client's time deposit. Therefore, after the customer's demand deposit account is transferred by the machine learning model 13, the balance in the account is still 5,000 yuan, which is between the upper limit of the difference and the lower limit of the difference.

In step 306, if the machine learning model 13 determines that the amount difference is greater than the upper limit of the amount difference, the machine learning model 13 generates an increased deposit amount greater than the fixed amount and transfers the increased deposit amount from the customer's current deposit account to the customer's time deposit account, and store the increased deposit amount and the corresponding transfer date in the external memory 122, wherein the increased deposit amount is the current fixed deposit and transfer amount that conforms to the current customer's financial income and expenditure status. For example, the customer initially sets the fixed deposit amount for each month to be 10,000 yuan, the minimum amount difference to be 2,000 yuan, and the upper limit to the amount difference to be 6,000 yuan. When the deposit amount in the customer's current deposit account is 20,000 yuan, the machine learning model 13 calculates that the difference is 10,000 yuan and determines that the difference is higher than the preset upper limit of the difference, then the machine learning model 13 changes the fixed deposit amount from 10,000 yuan to 10,000 yuan. Raised to 14,000 yuan, the additional deposit amount is 14,000 yuan. Therefore, after the customer's current deposit account is transferred through the machine learning model 13, the balance in the current deposit account will be 6,000 yuan, which will not exceed the upper limit of the difference.

In step 307, if the machine learning model 13 determines that the amount difference is less than zero, which means that the deposit amount in the customer's current deposit account is lower than the preset fixed amount, the machine learning model 13 does not transfer the account. In other embodiments, when the machine learning model 13 does not transfer funds, it further sends a short message to the mobile device of the customer to notify the customer that the balance in the account is insufficient.

FIG. 4 is a schematic diagram of an embodiment of a machine learning model of the automatic savings system of FIG. 2 . As shown in FIG. 4 , the neural network as the machine learning model 13 includes a first node N1 to a fourth node N4, wherein the first node N1 and the second node N2 are input nodes and the functions are [|( First input data)-(second input data)|-(third input data)], which means that the absolute value of the first input data minus the second input data is subtracted from the third input data. The three input data of the first node N1 are the deposit amount, the fixed amount and the upper limit of the amount difference respectively, wherein the deposit amount, the fixed amount and the upper limit of the amount difference are the first to third input data of the first node N1 respectively. The three input data of the second node N2 are the deposit amount, the fixed amount and the lower limit of the difference, wherein the deposit amount, the fixed amount and the lower limit of the difference are the first to third input data of the second node N2 respectively. As for the difference between the deposit amount and the fixed amount, the absolute value is the amount difference.

The third node N3 is an intermediate layer node and its function is to determine whether [amount difference - amount difference upper limit] and [amount difference - amount difference lower limit] are greater than 0, where the amount difference is the difference between the deposit amount and the fixed amount. The fourth node N4 is the output node and its function is the sum of the three input data, and the three input data of the fourth node N4 are (amount difference - amount difference upper limit) * first weight W1, (amount difference - amount difference lower limit)*second weight W2, and (fixed amount)*third weight W3. The output of the fourth node N4 of the neural network is the current fixed deposit transfer amount of the current month.

In this embodiment, the neural network has an online learning mechanism. If the output of the fourth node N4 of the neural network is smaller than the historical fixed deposit and transfer amount in the same month of the previous year, for example, the current amount in January 2020 If the fixed deposit and transfer amount is less than the historical fixed deposit and transfer amount in January 2019, the neural network is based on the fixed deposit and transfer amount between the current fixed deposit and transfer amount of the current month and the historical fixed deposit and transfer amount of the same month in the previous year. and through the backward pass algorithm, the first weight W1 is increased, the second weight W2 is increased, and the third weight W3 is decreased, wherein the adjustment percentage of the first weight W1 is less than the adjustment percentage of the second weight W2 and the third weight W3. For example, the adjustment percentage of increasing the first weight W1 is 12%, the adjustment percentage of increasing the second weight W2 is 20%, and the adjustment percentage of decreasing the third weight W3 is 20%. The above adjustment percentage is an example , not limited to this.

If the difference between the current fixed deposit and transfer amount of the neural network in the current year and the historical fixed deposit and transfer amount in the same month of the previous year is within a predetermined range, the neural network will maintain the current first A weight W1, a second weight W2 and a third weight W3.

If the current fixed deposit and transfer amount of the neural network in the current year is greater than the historical fixed deposit and transfer amount in the same month of the previous year, the neural network will base on the current fixed deposit and transfer amount of the current year and the history of the same month in the previous year. The difference between the fixed deposit and transfer amounts and the backward pass algorithm is used to increase the first weight W1, increase the second weight W2 and increase the third weight W3, wherein the adjustment percentage of the first weight W1 is greater than The adjustment percentage ratio between the second weight W2 and the third weight W3, for example, the adjustment percentage of the first weight W1 is increased by 20%, the adjustment percentage of the second weight W2 is increased by 12%, and the adjustment percentage of the third weight W3 is increased. The percentage is 12%, and the above percentage is an example, not limited thereto.

FIG. 5 is a flowchart of an automatic saving method according to a second embodiment of the present invention. As shown in FIG. 5 , the difference between the automatic saving method of the second embodiment and the automatic saving method of the first embodiment is that it further includes the interest rate management host 20 judging whether the average value of the fixed deposit and transfer amount of the account within a predetermined period is greater than the predetermined amount. The fixed amount is set. If the average value is greater than the fixed amount, the fixed deposit interest rate of the account will be increased.

In other embodiments, before the account monitoring host 10 of the automatic savings system collects the deposit data of the account in each predetermined period, the machine learning model 13 of the automatic savings system may also undergo offline learning, taking a neural network as an example based on the collected data. The historical deposit data (including the deposit amount and the date of the deposit) and the historical transfer data (including the transfer amount and the transfer date) of the account of the The difference upper limit and the amount difference lower limit are used for offline training, so that the fixed deposit and transfer amount output by the machine learning model 13 can be consistent with the personal financial transaction behavior of each account.

In this way, the more historical financial transaction data collected, the fixed deposit transfer calculated by the neural network after offline training when it receives the real-time deposit data, fixed amount, lower limit of amount difference, and upper limit of amount difference. The amount will also be more in line with the customer's income and expenditure situation, so as to achieve the effect of personal planning and arrangement of fixed deposit and transfer. Furthermore, when the external memory 122 stores the data of the fixed deposit transfer amount of each month of the previous year, the neural network can also adjust the weight according to the fixed deposit amount curve of the previous year.

In other embodiments, the automatic savings system can analyze the customer's transaction details over the past years. When the annual premium, tuition payment month or mortgage balance has been decreasing, there are more funds to save, which can be saved when the customer logs in to the automatic savings system. A prompt message will appear, reminding customers whether to modify the setting conditions of automatic savings (fixed amount, upper limit of amount difference and lower limit of amount difference) and display the current fixed savings plan.

In other embodiments, after calculating the fixed deposit and transfer amount, the automatic savings system does not perform the fixed deposit and transfer by itself, but can first output a confirmation operation page to the user's personal computer or mobile communication device, and the confirmation operation interface will display The current fixed deposit and transfer amount calculated by the automatic savings system is firstly filled in the input column of the fixed deposit and transfer amount on the confirmation operation page.

There is a confirmation button and a cancel button on the confirmation operation page. When the customer confirms that the current fixed deposit and transfer amount in the input column of the fixed deposit and transfer amount is correct, no additional adjustment is required, and the customer only needs to press the confirm button, then The automatic savings system conducts fixed deposit transfers. If the customer thinks that the current fixed deposit and transfer amount in the input column of the fixed deposit and transfer amount needs to be adjusted, he can also press the cancel button, and then enter the new current fixed deposit and transfer amount in the input column of the fixed deposit and transfer amount by himself, and then press the confirm button. Then the automatic savings system will transfer the fixed deposit according to the newly input current fixed deposit and transfer amount. In this way, the operation process of time deposit transfer can be greatly simplified, so as to reduce the occurrence of wrong time deposit amount transferred by the automatic savings system.

Based on the above, the automatic savings system and automatic savings method disclosed in the present invention can plan a customized fixed deposit and transfer amount and perform automatic fixed deposit and transfer according to the customer's actual personal financial income and expenditure status. When the customer has a high demand for funds, the automatic savings system automatically reduces the customer's fixed deposit and transfer amount. On the contrary, when the customer's funds are relatively abundant, the automatic savings system automatically increases the customer's fixed deposit amount, so the customer does not need to adjust it. Fixed deposit transfer amount. To achieve the effect of simplifying the operation process.

Although the present invention is disclosed in the foregoing embodiments, it is not intended to limit the present invention. Changes and modifications made without departing from the spirit and scope of the present invention belong to the scope of patent protection of the present invention. For the protection scope defined by the present invention, please refer to the attached patent application scope.

100: Automatic Savings System 10: Account Monitoring Host 12: Transfer management host 121: Processor 122: External memory 123: Internal memory 124: Drawing Circuits 13: Machine Learning Models 200: Automatic Savings System 20: Rate Management Host S: Bank internal server N1~N4: The first node ~ the fourth node W1~W3: The first weight ~ the third weight

FIG. 1 is a functional block diagram of an automatic saving system according to a first embodiment of the present invention. FIG. 2 is a functional block diagram of an automatic saving system according to a second embodiment of the present invention. FIG. 3 is a flowchart of an automatic saving method according to the first embodiment of the present invention. FIG. 4 is a schematic diagram of an embodiment of the machine learning model of the automatic savings system of FIG. 2 FIG. 5 is a flowchart of an automatic saving method according to a second embodiment of the present invention.

Claims (3)

An automatic deposit method using a neural network is implemented by an automatic deposit system. The automatic deposit method includes: collecting an existing deposit data of an account in each predetermined period with an account monitoring host of the automatic deposit system, the automatic deposit method The existing deposit data includes a deposit amount and a deposit time; a transfer management host of the automatic savings system executes a machine learning model, the machine learning model is a type of neural network, and a fixed amount, a The upper limit of the amount difference, a lower limit of the amount difference, and the deposit amount; the amount difference between the deposit amount and a fixed amount is calculated by this type of neural network; the difference between the amount and an amount difference lower limit and an amount are respectively calculated by the type of neural network If the neural network determines that the amount difference is greater than zero and less than or equal to the lower limit of the difference, the neural network outputs a current fixed deposit and transfer amount, and the current fixed deposit and transfer amount is less than the fixed amount. Amount; if this type of neural network determines that the amount difference is greater than the upper limit of the amount difference, use this type of neural network to output the current fixed deposit and transfer amount, and the current fixed deposit and transfer amount is greater than the fixed amount; if the current fixed deposit and transfer amount When the difference between the fixed deposit and transfer amount between the amount and a historical fixed deposit and transfer amount of the account is less than 0, the neural network is used to increase the number of the fixed deposit and transfer amount according to the difference of the fixed deposit and transfer amount and through a backward pass algorithm. a weight, increasing the second weight and decreasing the third weight, wherein a first adjustment percentage of the first weight is smaller than a second adjustment percentage of the second weight and a third adjustment percentage of the third weight ; If the difference between the current fixed deposit and transfer amount and the historical fixed deposit and transfer amount falls within a predetermined amount range, the neural network maintains the first weight, the second weight and the third weight; if the difference between the fixed deposit and transfer amount is greater than 0, the neural network is used to increase the first weight and increase the third weight according to the difference between the fixed deposit and transfer amount and through the backward pass algorithm. Second weight and increase the third weight, wherein the first adjustment percentage is greater than the second adjustment percentage and the third adjustment percentage; if the type of neural network determines that the amount difference is less than 0, the type of neural network does not perform Transfer money; use a drawing circuit of the transfer management host to generate a current year's fixed deposit amount curve according to a current fixed deposit transfer amount for a whole year, and transmit the fixed deposit amount curve to an electronic device of a customer through a network mailbox; and receiving a password to reset the fixed amount, the upper limit of the difference and the lower limit of the difference. The automatic saving method using a neural network as described in claim 1, wherein the predetermined period is one month. The automatic saving method using a neural network as described in claim 1, further comprising judging whether the average value of the fixed deposit and transfer amount of the account in a predetermined period is greater than the fixed amount, and if it is greater than the fixed amount, increasing the account's balance Fixed deposit interest rate.

Images