TWM634647U - rate estimator - Google Patents

Abstract

The present invention proposes a rate estimating device, which is installed in a bank host server, wherein the rate estimating device is connected to a database and a user device, and the database stores at least one historical rate information and at least one historical sales volume information. The rate estimation device includes a data analysis module, a data integration module and a machine learning calculation module. The data analysis module is used to generate statistical reference revenue information based on historical rate information and historical sales volume information. The data integration module is used to receive the statistical reference revenue information and generate a calculation instruction information. The machine learning calculation module is used to receive calculation instruction information, use the analysis method to generate a fee revenue formula, and generate suggested fee rate information based on the fee revenue formula.

Description

rate estimator

The present invention relates to a rate estimating device, in particular to a rate estimating device which can automatically calculate a suggested rate by using an analysis method according to historical rate information and historical sales information.

Banks often offer various preferential activities (for example, interest rate, exchange rate, handling fee discounts) to increase sales and maintain customer loyalty to the bank. However, if the preferential quota is poorly set, it will affect the bank's profit, and the promotional activities launched will become a stumbling block to the bank's profit. If you set too much preferential quota, you may not be able to effectively increase the sales volume of customers, and if you set too few discounts, you will not be able to increase the sales volume of customers.

In this regard, how to automatically estimate the forecast and dynamically adjust the parameter value of the preferential rate, use the big data of the past history to analyze the bank's past preferential rate and sales volume data, and generate the optimal preferential rate value, which is the technology issues to be explored in the field.

Therefore, in order to solve the problems in the prior art, the embodiment of the present invention proposes a tariff estimation device. The rate estimating device is signally connected to the database and the user device, and the rate estimating device includes a data analysis module, a data integration module and a machine learning calculation module. Wherein, the above-mentioned database stores at least one piece of historical rate information and at least one piece of historical sales volume information.

The signal of the data analysis module is connected to the above database. The data analysis module generates statistical reference revenue information based on the above-mentioned historical rate information and the above-mentioned historical sales volume information.

The data integration module is connected to the data analysis module by signal. The data integration module receives statistical reference revenue information and generates calculation instruction messages.

The signal of the machine learning computing module is connected to the data integration module. The machine learning calculation module receives the calculation instruction information, and uses the analysis method to analyze the corresponding relationship between historical rate information and historical sales volume information to generate a fee revenue formula, and generates suggested rate information based on the fee revenue formula.

According to some embodiments of the present invention, the tariff estimating device further includes a data output module, the signal of which is connected to the machine learning calculation module. The above-mentioned data output module outputs the above-mentioned suggested rate information to the above-mentioned user device, and outputs the final rate information and sends it back to the above-mentioned database.

According to some embodiments of the present invention, the above-mentioned historical rate information and the above-mentioned historical sales volume information include multiple businesses, and the multiple business lines include items with preferential rates and items without preferential rates.

According to some embodiments of the present invention, the above-mentioned multiple businesses include live storage, credit granting, foreign exchange, funds, insurance or any combination of the foregoing.

According to some embodiments of the present invention, the above-mentioned statistical reference revenue information records multiple reference fee rates and multiple reference sales volumes corresponding to the reference fee rates.

According to some embodiments of the present invention, the above-mentioned final rate information is the usage rate information generated by the above-mentioned user device or the above-mentioned suggested rate information.

According to some embodiments of the present invention, the above analysis method is maximum likelihood estimation method, linear regression, logistic regression, k-nearest neighbor algorithm or decision tree.

According to some embodiments of the present invention, the above-mentioned service fee revenue formula includes the original fee rate, the preferential fee rate and the sales volume. The difference between the above-mentioned original fee rate and the above-mentioned preferential fee rate is multiplied by the above-mentioned sales volume to generate service fee revenue.

According to some embodiments of the present invention, the above-mentioned suggested fee rate information includes the suggested preferential fee rate and the maximum handling fee revenue generated corresponding to the above-mentioned suggested preferential fee rate. The above fee revenue formula corresponds to the above maximum fee revenue.

According to some embodiments of the present invention, the tariff estimating device further includes an error detection module, the signal of which is connected to the machine learning calculation module and the user device. The above-mentioned error detection module checks whether the difference between the above-mentioned original rate and the above-mentioned suggested preferential rate exceeds a threshold, and if it exceeds the above-mentioned threshold, a warning message is sent to the above-mentioned user device.

In order to make the purpose, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present invention but not to limit the present invention.

Please refer to FIG. 1 . FIG. 1 is a block diagram of a tariff estimation device 100 according to an embodiment of the present invention. In FIG. 1 , the rate estimation device 100 is connected to an external database 110 and a user device 111 (for example, a computer, a tablet, an electronic device), wherein the above-mentioned database 110 (for example, data inside or outside the bank database) to store at least one piece of historical rate information and at least one piece of historical sales volume information. The rate estimation device 100 includes a data analysis module 101 , a data integration module 102 , a machine learning calculation module 103 , a data output module 104 , and an error detection module 105 .

In FIG. 1 , the data analysis module 101 is connected to the database 110 by signals. The above-mentioned data analysis module 101 generates statistical reference revenue information according to the above-mentioned historical rate information and the above-mentioned historical sales volume information. For example, taking each year as a data volume, the rate used in each month of the year and the sales volume generated by the corresponding rate are arranged and recorded in a table.

In one embodiment, the above-mentioned statistical reference revenue information records multiple reference fee rates and multiple reference sales volumes corresponding to the reference fee rates.

In one embodiment, the above-mentioned historical rate information and the above-mentioned historical sales volume information include multiple businesses, and the multiple business lines include items with preferential rates and items without preferential rates. For example, not limited to this model, the above reference revenue information includes items with preferential rates and items without preferential rates, and the corresponding individual sales volume, so as to analyze whether there are preferential rates for specific businesses Is there any significant difference in sales volume in the rate.

In one embodiment, the above-mentioned multiple businesses are survival, credit, foreign exchange, fund, insurance or any combination of the aforementioned types.

The data integration module 102 is connected to the data analysis module 101 by signal. The above-mentioned data integration module 102 receives statistical reference revenue information, generates a calculation command message, and sends it to the machine learning calculation module 103, requiring the machine learning calculation module 103 to start automatic analysis based on the statistical reference revenue information.

The machine learning calculation module 103 is connected to the data integration module 102 by signal. The machine learning calculation module 103 receives the calculation instruction information, and uses the analysis method to analyze the corresponding relationship between the historical rate information and the historical sales volume information to generate a fee revenue formula, and generates suggested fee rate information according to the fee revenue formula.

。分析出來的結果為q=1+y。 For example, not limited to the present invention, the analysis method uses statistical regression analysis to analyze the relationship between the premium rate (y) and the sales volume (q). It can be known from the simple regression formula that sales volume (q) = first parameter (a) + second parameter (B) * preferential rate (y). The second parameter (B) is calculated by the maximum likelihood estimation method (MLE method). first parameter

. The analyzed result is q=1+y.

In one embodiment, the above analysis method is the maximum likelihood estimation method, linear regression, logistic regression, k-nearest neighbor algorithm or decision tree.

In an implementation manner, the above-mentioned service fee revenue formula includes the original fee rate, the preferential fee rate and the sales volume. The difference between the above-mentioned original fee rate and the above-mentioned preferential fee rate is multiplied by the above-mentioned sales volume to generate service fee revenue.

For example, not limited to this model, service fee revenue (z) = (original fee rate - preferential fee rate y) * sales volume (q). For example, if the original fee rate is 10, z=(10-y)*q, substituting q=1+y in the above example, z=-y ² +9y+10 can be obtained. When the preferential fee rate (y) is 4.5, the service fee revenue (z) has a maximum value of 30.25.

In one embodiment, the above-mentioned suggested fee rate information includes the suggested preferential fee rate and the maximum service fee revenue generated corresponding to the above-mentioned suggested preferential fee rate. The fee revenue formula corresponds to the above-mentioned maximum fee revenue.

In another embodiment, the above-mentioned data output module 104 is connected to the machine learning computing module 103 by signal. The data output module 104 outputs the suggested rate information to the user device 111 , and generates the final rate information and sends it back to the above-mentioned database 110 .

In yet another implementation manner, the above-mentioned final rate information is the usage rate information generated by the above-mentioned user device 111 or the above-mentioned suggested rate information.

In another embodiment, the above-mentioned debugging module 105 is signally connected to the machine learning computing module 103 and the user device 111 . The error detection module 105 checks whether the difference between the above-mentioned original rate and the above-mentioned preferential rate exceeds a threshold, and if it exceeds the above-mentioned threshold, then sends a warning message to the above-mentioned user device 111 . For example, not limited to the present model, the threshold value is 0.05. When the difference between the originally running rate and the suggested preferential rate exceeds 0.05, the debugging module 105 determines that the rate estimation system may be abnormal.

Please refer to FIG. 2 . FIG. 2 is a flow chart of the rate estimation step according to an embodiment of the present invention. The method based on rate estimation can be executed on the rate estimation device 100 and includes the following steps 201 to 207 .

In step 201, the data analysis module 101 downloads historical rate information and historical sales volume information from the database 110 to generate statistical reference revenue information.

In step 202, the data integration module 102 generates a calculation instruction message according to the above statistical reference revenue information.

In step 203, the machine learning calculation module 103 receives the above-mentioned calculation instruction message, and uses an analysis method to analyze the corresponding relationship between the historical rate information and the historical sales volume information to generate a commission revenue formula.

In step 204 , generate suggested fee rate information according to the fee income formula, and output the suggested fee rate information to the user device 111 .

In step 205, the data output module 104 determines whether to use the suggested rate information, if not, proceed to step 206, and if use the suggested rate information, then proceed to step 207.

In step 206 , the user device 111 generates usage rate information, and the data output module 104 returns the usage rate information to the database 110 .

In step 207 , the data output module 104 returns the suggested rate information to the database 110 .

(only one paragraph number) 100: rate estimation device 101:Data analysis module 102: Data integration module 103:Machine Learning Computing Module 104: Data output module 105: Debugging Module 110: Database 111: user device 201~207: Steps

FIG. 1 is a functional block diagram of a tariff estimation device according to an embodiment of the present invention. FIG. 2 is a method flow chart of the rate estimation step according to an embodiment of the present invention.

100: rate estimation device

101:Data analysis module

102: Data integration module

103:Machine Learning Computing Module

104: Data output module

105: Debugging Module

110: Database

111: user device

Claims (10)

A rate estimating device is installed in a bank host server, wherein the rate estimating device is connected to a database and a user device, and the database stores at least one historical rate information and at least one historical Sales volume information, including: A data analysis module, signally connected to the database, for generating statistical reference revenue information based on the historical rate information and the historical sales volume information; A data integration module, signally connected to the data analysis module, used to receive the statistical reference revenue information and generate a calculation instruction message; and A machine learning calculation module, signally connected to the data integration module, used to receive the calculation instruction information, and use an analysis method to analyze the corresponding relationship between the historical rate information and the historical sales volume information to generate a A service fee revenue formula, and generate a suggested rate information according to the service fee revenue formula. The rate estimating device as described in claim 1, further comprising: A data output module, signal-connected to the machine learning calculation module for outputting the suggested rate information to the user device and outputting a final rate information back to the database. The rate estimating device according to claim 1, wherein the historical rate information and the historical sales volume information include a plurality of businesses, wherein the plurality of business lines include items with preferential rates and items without preferential rates . The rate estimating device according to claim 3, wherein the plurality of services are selected from live storage, credit, foreign exchange, fund, insurance or any combination of the foregoing. The fee rate estimating device according to claim 1, wherein the statistical reference revenue information records multiple reference fee rates and multiple reference sales volumes corresponding to the reference fee rates. The rate estimating device according to claim 1, wherein the final rate information is a usage rate information generated by the user device or the suggested rate information. The rate estimating device according to claim 1, wherein the analysis method is the maximum likelihood estimation method, linear regression, logistic regression, k-nearest neighbor algorithm or decision tree. The fee rate estimating device as described in Claim 1, wherein the service fee revenue formula includes an original fee rate, a preferential fee rate and a sales volume, and the difference between the original fee rate and the preferential fee rate Multiplying the sales volume generates a commission revenue. The fee rate estimating device as described in claim 8, wherein the suggested fee rate information includes a suggested preferential fee rate and a maximum commission revenue generated corresponding to the suggested preferential fee rate, when the preferential fee rate is If the recommended preferential fee rate is selected, the preferential fee rate will generate the maximum commission revenue according to the commission revenue formula. The rate estimating device as described in Claim 8, further comprising: An error detection module, signally connected to the machine learning calculation module and the user device, to check whether the difference between the original tariff rate and the suggested discounted tariff rate exceeds a threshold, and if the difference exceeds the threshold , an alert message is sent to the user's device.

Images