Method and device for determining banknote adding suggestion at bank teller terminal

Abstract

The present invention provides a method and device for determining bank teller terminal cash addition suggestions, which relate to the financial field and the artificial intelligence field. The method comprises: obtaining terminal device attribute data, cash addition personnel information and preset cash addition plan data of a bank teller terminal to be added with cash; determining terminal device cash addition feature data according to the terminal device attribute data and cash addition personnel information data; determining terminal cash addition plan suggestion data according to the terminal device cash addition feature data, preset cash addition plan data and a pre-trained cash addition suggestion model; wherein the cash addition suggestion model is a model established by machine learning based on the pre-acquired terminal feature data and cash addition personnel feature data. In the process of designing the cash addition plan, auxiliary equipment management evaluation suggestions are given to business personnel, guiding business personnel to optimize the ATM cash addition plan, improving the accuracy of the cash addition plan, and directly improving the equipment management level of business personnel.

Description

Technical Field

[0001] The invention relates to artificial intelligence technology, in particular to a method and device for determining banknote adding suggestion at a bank teller terminal. Background Art

[0002] Traditional banks' ATM equipment operations mainly adopt a management model that manually evaluates the cash usage of ATM equipment and manually designs cash replenishment plans. When the ATM equipment needs to be replenished and how much cash should be replenished, these questions rely on the expert experience of ATM equipment management staff to make assessments and decisions.

[0003] With the development of artificial intelligence, the operation of ATM equipment in banks has gradually become more refined, automated and intelligent. Intelligent algorithms are introduced to predict the cash usage of ATM equipment and evaluate the priority of ATM equipment recharging, further providing additional data support for business personnel's ATM equipment operation and management. The existing ATM equipment operation system uses intelligent algorithms to help business personnel understand the current status of ATM equipment more intuitively, and indirectly improve the equipment management level of business personnel. However, in the existing technology, in the design process of the actual recharging plan, business personnel may still make wrong decisions with large errors from the actual cash demand of ATM equipment due to lack of personal experience and unclear understanding of the characteristics of ATM equipment, resulting in evaluation errors. Therefore, it is of great significance to establish a recharging plan auxiliary system to improve the design level of business personnel. Summary of the invention

[0004] In view of the shortcomings of the prior art, an embodiment of the present invention provides a method for determining a banknote adding suggestion at a bank teller terminal, comprising:

[0005] Obtain terminal device attribute data of the bank teller terminal to be loaded with banknotes, information about the banknote loading personnel, and preset banknote loading plan data;

[0006] Determine the terminal device adding banknote characteristic data according to the terminal device attribute data and the banknote adding personnel information data;

[0007] The terminal adding banknotes plan suggestion data is determined according to the terminal equipment adding banknotes characteristic data, the preset adding banknotes plan data and the pre-trained adding banknotes suggestion model; wherein the adding banknotes suggestion model is a model established by machine learning based on the pre-acquired terminal characteristic data and the adding banknotes personnel characteristic data.

[0008] In the embodiment of the present invention, the terminal device attribute data includes: terminal transaction data, weather data of the location of the terminal device, banknote addition date, payday and repayment date data;

[0009] The information of the cash adding personnel includes: identity information, working years and professional information of the cash adding personnel.

[0010] In the embodiment of the present invention, the terminal equipment cash adding characteristic data includes: basic characteristic data of the terminal equipment, basic characteristic data of the cash adding personnel, clustering characteristic data and expert characteristic data; wherein,

[0011] The method of determining the terminal device adding banknote characteristic data according to the terminal device attribute data and the banknote adding personnel information data includes:

[0012] Performing feature construction on the terminal device attribute data and the cash adding personnel information data to determine the basic feature data of the terminal device and the basic feature data of the cash adding personnel;

[0013] Performing clustering processing on the basic characteristic data of the terminal device and the basic characteristic data of the cash adding personnel to generate cluster characteristic data;

[0014] The expert feature data is determined according to the preset expert rules and cluster feature data.

[0015] In the embodiment of the present invention, the method further includes: performing machine learning to establish a banknote adding suggestion model based on the historical terminal device banknote adding feature data and historical banknote adding plan data acquired in advance; which includes:

[0016] The previously acquired historical terminal device note adding feature data is used as input features, and the historical note adding plan data is used as prediction labels to train a candidate machine learning model of preset hyperparameters;

[0017] Determine a preferred machine learning model from the candidate machine learning models according to the model training results;

[0018] The banknote adding suggestion model is determined according to the preset weight setting and the determined preferred machine learning model.

[0019] In the embodiment of the present invention, the method further includes:

[0020] If it is determined that the training time of a single model in the preferred machine learning model is greater than a preset time threshold and / or the prediction results of a single algorithm model have abnormal deviations for more than a preset number of consecutive days, the single algorithm model will be removed from the preferred machine learning model.

[0021] At the same time, the present invention also provides a bank teller terminal adding banknote suggestion determining device, comprising:

[0022] An interactive module, used to obtain terminal device attribute data of the bank teller terminal to be loaded with banknotes, information about the banknote loading personnel, and preset banknote loading plan data;

[0023] A data processing module, used to determine the terminal device adding banknote characteristic data according to the terminal device attribute data and the banknote adding personnel information data;

[0024] The recommendation data generation module is used to determine the terminal banknote adding plan recommendation data based on the terminal equipment banknote adding characteristic data, the preset banknote adding plan data and the pre-trained banknote adding recommendation model; wherein the banknote adding recommendation model is a model established by machine learning based on the pre-acquired terminal characteristic data and banknote adding personnel characteristic data.

[0025] In the embodiment of the present invention, the terminal equipment cash adding characteristic data includes: basic characteristic data of the terminal equipment, basic characteristic data of the cash adding personnel, clustering characteristic data and expert characteristic data; wherein,

[0026] The data processing module comprises:

[0027] A preprocessing unit, used for constructing features of the terminal device attribute data and the banknote adding personnel information data to determine the basic feature data of the terminal device and the basic feature data of the banknote adding personnel;

[0028] A clustering unit, used for clustering the basic characteristic data of the terminal device and the basic characteristic data of the cash adding personnel to generate cluster characteristic data;

[0029] The rule processing unit is used to determine the expert feature data according to the preset expert rules and the cluster feature data.

[0030] In an embodiment of the present invention, the device further comprises: a model training module, which is used to perform machine learning to establish a banknote adding suggestion model based on the previously acquired historical terminal device banknote adding feature data and historical banknote adding plan data; the model training module comprises:

[0031] A model training unit, used to train a candidate machine learning model of preset hyperparameters by using the previously acquired historical terminal device note adding feature data as input features and the historical note adding plan data as prediction labels;

[0032] A preferred model unit, used to determine a preferred machine learning model from the candidate machine learning models according to the model training result;

[0033] The weight optimization unit is used to determine the banknote adding suggestion model according to the preset weight setting and the determined preferred machine learning model.

[0034] In an embodiment of the present invention, the device further comprises:

[0035] An updating module is used to determine that if the training time of a single model in the preferred machine learning model is greater than a preset time threshold and / or the prediction results of a single algorithm model have abnormal deviations for more than a preset number of consecutive days, the single algorithm model will be removed from the preferred machine learning model.

[0036] At the same time, the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the above method is implemented when the processor executes the computer program.

[0037] At the same time, the present invention also provides a computer-readable storage medium, which stores a computer program for executing the above method.

[0038] The present invention provides a method and device for determining bank teller terminal cash addition suggestions, which determine terminal equipment cash addition feature data according to acquired terminal equipment attribute data of a bank teller terminal to be added with cash, cash addition personnel information and preset cash addition plan data; determine terminal cash addition plan suggestion data according to the terminal equipment cash addition feature data, preset cash addition plan data and a pre-trained cash addition suggestion model, that is, based on a machine learning algorithm, comprehensively use data analysis, supervised learning, clustering and other technologies, for different categories of ATM equipment and different business personnel, fit the difference between the ATM cash addition plan designed by the business personnel in history and the actual consumption amount, give auxiliary equipment management evaluation suggestions to the business personnel in the design process of the cash addition plan, guide the business personnel to optimize the ATM cash addition plan, improve the accuracy of the cash addition plan, and directly improve the equipment management level of the business personnel.

[0039] In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0041] Figure 1 A flow chart of a method for determining banknote adding suggestion for a bank teller terminal provided by the present invention;

[0042] Figure 2 It is an overall structural diagram of an ATM equipment banknote adding design auxiliary system of the present invention;

[0043] Figure 3 It is a processing flow chart of a preprocessing module of an ATM equipment banknote adding design auxiliary system of the present invention;

[0044] Figure 4 It is a block diagram of a cluster analysis module of an ATM equipment banknote adding design auxiliary system of the present invention;

[0045] Figure 5 It is a block diagram of a prediction algorithm module of an ATM equipment banknote adding design auxiliary system of the present invention;

[0046] Figure 6 It is an update strategy diagram of a self-update module of an ATM device banknote adding design auxiliary system of the present invention;

[0047] Figure 7 A block diagram of a device for determining banknote adding suggestion for a bank teller terminal provided by the present invention;

[0048] Figure 8 A schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0049] 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.

[0050] In the prior art, during the design of the actual cash replenishment plan, business personnel may still make wrong decisions that are far from the actual cash demand of the ATM equipment due to lack of personal experience and unclear understanding of the characteristics of the ATM equipment, resulting in evaluation errors. Therefore, it is of great significance to establish a cash replenishment plan auxiliary system to improve the design level of business personnel.

[0051] like Figure 1 As shown, a bank teller terminal cash addition suggestion determination method provided by the present invention comprises:

[0052] Step S101, obtaining the terminal device attribute data of the bank teller terminal to be loaded with banknotes, the information of the banknote loading personnel and the preset banknote loading plan data;

[0053] Step S102, determining the terminal device adding banknote characteristic data according to the terminal device attribute data and the banknote adding personnel information data;

[0054] Step S103, determining the terminal adding banknote plan suggestion data according to the terminal equipment adding banknote characteristic data, the preset adding banknote plan data and the pre-trained adding banknote suggestion model; wherein the adding banknote suggestion model is a model established by machine learning based on the pre-acquired terminal characteristic data and the adding banknote personnel characteristic data.

[0055] The present invention assists business personnel in equipment management through intelligent algorithms, and realizes an ATM equipment adding banknote design assistance. The present invention automatically gives equipment management evaluation suggestions, directly improves the equipment management level of business personnel, and provides a reference for the application of intelligent algorithms in the field of ATM equipment operation.

[0056] In the embodiment of the present invention, the terminal device attribute data includes: terminal transaction data, weather data of the location of the terminal device, banknote addition date, payday and repayment date data;

[0057] The information of the cash adding personnel includes: identity information, years of work experience and profession of the cash adding personnel.

[0058] The bank teller terminal cash addition suggestion determination method provided by the present invention adopts a machine learning algorithm to fit the historical cash addition plan design errors of the business personnel according to the terminal device attribute data and the cash addition personnel information, aiming at different types of ATM devices and different business personnel, so as to provide automatic evaluation and optimization suggestions for the cash addition plan designed by the business personnel, and guide the business personnel to optimize the ATM cash addition plan.

[0059] In the embodiment of the present invention, the terminal equipment cash adding characteristic data includes: basic characteristic data of the terminal equipment, basic characteristic data of the cash adding personnel, clustering characteristic data and expert characteristic data; wherein,

[0060] The method of determining the terminal device adding banknote characteristic data according to the terminal device attribute data and the banknote adding personnel information data includes:

[0061] Performing feature construction on the terminal device attribute data and the cash adding personnel information data to determine the basic feature data of the terminal device and the basic feature data of the cash adding personnel;

[0062] Performing clustering processing on the basic characteristic data of the terminal device and the basic characteristic data of the cash adding personnel to generate cluster characteristic data;

[0063] The expert feature data is determined according to the preset expert rules and cluster feature data.

[0064] Specifically, in the embodiment of the present invention, the basic characteristic data of the terminal device and the basic characteristic data of the cash adding personnel are determined through the following steps:

[0065] The collected data will be analyzed one by one according to each dimension. The data conditions analyzed include but are not limited to the missing ratio common to all features, the numerical range and numerical magnitude of numerical features.

[0066] Different strategies will be adopted for data in different conditions to carry out data cleaning, data integration, data reduction and data transformation to improve data quality.

[0067] The traditional feature construction method will be used to construct appropriate features. Specifically, the monthly data will be regarded as category features and 12 columns of features will be constructed.

[0068] The list of high-importance feature combinations obtained in advance will be used as feature supplement combination features. Accurate, complete, and high-quality feature data should be obtained as much as possible. In this embodiment, the list of high-importance feature combinations is obtained by repeatedly training the model offline, based on the comprehensive evaluation of the model effect (RMSE, the root mean square error between the predicted value and the true value) and the importance of the feature. The algorithms used for the evaluation include but are not limited to traditional machine learning algorithms such as random forests and linear regression. The feature combination with the highest evaluation importance will be used as the list of high-importance feature combinations.

[0069] Specifically, in the embodiment of the present invention, clustering processing is performed on the basic characteristic data of the terminal device and the basic characteristic data of the cash adding personnel to generate cluster characteristic data, and the expert characteristic data is determined according to the preset expert rules and the cluster characteristic data, including:

[0070] It should be noted that the clustering algorithms in this embodiment include but are not limited to common clustering algorithms such as K-Means, EM, and DBSCAN.

[0071] This embodiment uses the K-Means algorithm and the basic features of the ATM devices as input to reduce the categories of the ATM devices to a number that the business experts consider to be more appropriate, and determines whether the number of categories obtained by clustering the ATM devices is at an appropriate level. Otherwise, cluster analysis is performed again.

[0072] Based on the ATM categories, the business experience of business experts is introduced, and by loading the pre-sorted expert rules, corresponding expert features are generated for each category of ATM equipment, including but not limited to the priority of ATM equipment adding cash, the degree of influence by payday, the degree of influence by holidays, etc. The intervention of expert rules will provide the model with higher-level modeling information than the basic data features.

[0073] In the embodiment of the present invention, the method further includes: performing machine learning to establish a banknote adding suggestion model based on the historical terminal device banknote adding feature data and historical banknote adding plan data acquired in advance; which includes:

[0074] The previously acquired historical terminal device note adding feature data is used as input features, and the historical note adding plan data is used as prediction labels to train a candidate machine learning model of preset hyperparameters;

[0075] Determine a preferred machine learning model from the candidate machine learning models according to the model training results;

[0076] The banknote adding suggestion model is determined according to the preset weight setting and the determined preferred machine learning model.

[0077] Specifically, in an embodiment of the present invention, a plurality of candidate algorithms (i.e., candidate machine learning models) are pre-set including but not limited to: SVM, KNN, RandomForest, GBDT, xgboost, LSTM, and each algorithm has its own unique one or more "hyperparameters".

[0078] The preset candidate algorithms constitute an algorithm pool, and each group (algorithm, hyperparameter) combination is an "optional algorithm". In this embodiment, the same feature input and prediction label will be used for each "optional algorithm" to train the model in conjunction with the preset hyperparameters. The single algorithm model training unit 41 will maximize the use of machine resources, and perform model training and effect evaluation of multiple "optional algorithms" in parallel according to the host machine resources.

[0079] Each algorithm has one or more unique "hyperparameters". For RandomForest, for example, the hyperparameters include the number of trees and the depth of the tree. For a certain type of data, after multiple model trainings, we can find several sets of hyperparameters that usually perform better (for example, the number of trees = 100, the depth = 20). These hyperparameters with better performance will be selected as the preset parameters and used in the model training of the single algorithm here.

[0080] The preferred model is selected from all single algorithm models according to preset rules. The selection method of this embodiment is divided into two steps. The first step is to select the single algorithm model with the best evaluation effect of each algorithm according to the algorithm. The second step is to select the top 4 models of the evaluation effect from the single algorithm models selected in the first step.

[0081] Provide appropriate weight settings for the preferred model to improve the prediction effect of the combined model. For example, there are four preferred models given. In the model prediction, the preferred model gives four prediction results, namely x1, x2, x3, x4, and the actual result is y. Take x1, x2, x3, x4 as features and y as a label to train a linear regression model and obtain the optimal weight values ​​w1, w2, w3, w4. Use the optimal weight vector to perform weighted averaging on the prediction values ​​of the preferred model to obtain the final combined prediction result.

[0082] In the embodiment of the present invention, the method further includes:

[0083] If it is determined that the training time of a single model in the preferred machine learning model is greater than a preset time threshold and / or the prediction results of a single algorithm model have abnormal deviations for more than a preset number of consecutive days, the single algorithm model will be removed from the preferred machine learning model.

[0084] Specifically, in the embodiment of the present invention, during the process of processing the data of the banknote addition plan suggestion, the prediction effect of the single algorithm model is evaluated to see whether it is stable. If the last training time of the single algorithm model reaches the preset maximum value or the prediction result of the single algorithm model has abnormal deviations for N consecutive days (specifically, the error between the predicted value and the true value is higher than 150% of the average error in the training data), the model is kicked out of the preferred model list and the single algorithm model is retrained. If the single algorithm model is self-updated or the prediction effect of the combined model is lower than the threshold set by the preset rules, the weight parameters of the combined model are adjusted, that is, the linear regression model is retrained.

[0085] The present invention provides an auxiliary method for designing cash addition of ATM equipment, which assists business personnel in giving equipment management evaluation suggestions during the design process of cash addition plan, improves the accuracy of cash addition plan, and directly improves the equipment management level of business personnel.

[0086] Specifically, an embodiment of the present invention provides an ATM equipment cash adding design auxiliary system and method, which utilizes machine learning clustering technology to classify ATM equipment, and adopts a machine learning algorithm to fit the historical cash adding plan design errors of the business personnel for different categories of ATM equipment and different business personnel, thereby providing automatic evaluation and optimization suggestions for the cash adding plans designed by the business personnel, and guiding the business personnel to optimize the ATM cash adding plans.

[0087] In this embodiment, the error in the banknote adding plan design is the difference between the ideal banknote adding amount and the planned banknote adding amount. The ideal banknote adding amount is equal to the actual cash consumption of the device from the current banknote adding to the next banknote adding. Since banks generally adopt the method of replacing the banknote box for banknote adding to the ATM device, when the banknote adding amount is 500,000, it means that the available amount after the ATM device adds banknotes is 500,000, and the balance of the banknote box before the ATM device adds banknotes will be taken back to the bank for counting.

[0088] Specifically, for example, an ATM device has two cash additions from January 5 to January 8, with the dates of the additions being January 5 and January 8 respectively. The amount of cash added on January 5 is A5, and the balance of the cash box withdrawn from the ATM is B5. The withdrawal amount from January 5 to January 8 is OUT, and the deposit amount is IN. The amount of cash added on January 8 is A8, and the balance withdrawn from the ATM is B8. Then:

[0089] The design error of the ATM equipment's note adding plan on January 5 is E=A5+IN-OUT;

[0090] In fact, the error of adding money is B8 = A5 + IN-OUT, so the actual error of adding money plan design is E = B8. Therefore, for an ATM device that is determined to add money in the ATM adding money plan, the error of adding money plan design refers to the difference between the actual cash consumption of the device after this adding money and before the next adding money and the amount of this adding money, which is equal to the balance of the cash box at the next adding money.

[0091] The technical solution of the present invention will be described in detail below in conjunction with the accompanying drawings of the present invention.

[0092] Figure 2 The overall structure diagram of an ATM device banknote adding design auxiliary system provided by an embodiment of the present invention is as follows: Figure 2 As shown, the present embodiment provides an ATM device adding banknote design auxiliary system comprising:

[0093] User interaction module 1, data preprocessing module 2, cluster analysis module 3, prediction algorithm module 4 and self-update module 5.

[0094] The user interaction module 1 is the interactive entry between the user and the ATM equipment cash addition design auxiliary system, and is embedded with the interactive function of the ATM cash addition plan design page. When the business personnel are designing the ATM cash addition plan and filling in the ATM cash addition plan on the page, the user interaction module 1 will interact with the prediction algorithm module 4, and transmit the basic information of the business personnel and the manually designed ATM cash addition plan to the prediction algorithm module 4, so as to obtain quasi-real-time ATM cash addition plan design suggestions, and assist the business personnel in optimizing the design of the ATM cash addition plan.

[0095] The pre-processing module 2, the data entry of the intelligent model of the ATM equipment cash addition design auxiliary system, will provide formatted data for the cluster analysis module 3 and the prediction algorithm module 4. The data required by the ATM equipment cash addition design auxiliary system is diverse, including not only the basic characteristic data of ATMs that affect the cash usage of ATMs, such as ATM transaction data, weather, holidays, and payday repayment dates, but also the basic characteristic data of the business personnel, such as the human resources number, years of work, and profession, and the cash addition amount designed by the business personnel.

[0096] In the embodiment of the present invention, after processing the banknote adding plan suggestion data, banknote adding plan data for a certain ATM machine is generated. The banknote adding designer designs to add 100,000 banknotes tomorrow. Through the input data model analysis and prediction, it may output a suggestion of "it is recommended to increase the amount by 50,000" and so on. The suggestion for adjusting the banknote adding amount guides the banknote adding designer to add banknotes.

[0097] In this embodiment, the detailed steps of the preprocessing module 2 are shown in Figure 3 shown.

[0098] Figure 3The figure shows the fluid flow diagram processed by the pre-processing module 2 of the ATM equipment banknote adding design auxiliary system in this embodiment, which specifically includes:

[0099] S201, a data quality analysis step according to dimensions, analyzes the collected data one by one according to each dimension, and the analyzed data conditions include but are not limited to the missing ratio common to all features, the numerical range and numerical magnitude of numerical features, etc.

[0100] In this embodiment, the data conditions analyzed include but are not limited to the missing ratio common to all features, the numerical range and numerical magnitude of numerical features, etc.

[0101] Specifically, in the embodiment of the present invention, when cleaning data, features with a missing ratio higher than a certain threshold will be combined with manual data importance analysis to decide whether to directly abandon the feature;

[0102] For features with particularly high data levels, some common machine learning processing methods can be used, such as taking log(x), where x is the value of the original feature, to reduce the order of magnitude of the feature and obtain higher quality features.

[0103] Step S202, data cleaning, adopts different strategies for data in different conditions to perform data cleaning, data integration, data reduction and data transformation to improve data quality.

[0104] Step S203, feature construction. In this embodiment, a traditional feature construction method is used to construct appropriate features. Specifically, the month data is regarded as a category feature to construct 12 columns of features.

[0105] In this embodiment, the features of the ATM device are mainly of three types:

[0106] First, attribute features, such as geographic location, are basically category features and need to be processed by category.

[0107] The second is the flow characteristics, which include the recent deposit and withdrawal volume, historical deposit and withdrawal volume, etc., which are numerical characteristics with size relationships. This part can be analyzed according to the actual characteristic effect, and can be classified according to the deposit and withdrawal range (10,000-50,000 is "small amount", 50,000-100,000 is "normal", etc.), or directly treated as a numerical value

[0108] The third is the features given based on preset expert experience. This part is generally regarded as a numerical value with a size relationship.

[0109] Step S204, finally, the combined features are constructed, and the list of highly important feature combinations obtained in advance is used as a feature supplement combined feature to obtain accurate, complete and high-quality feature data as much as possible.

[0110] Among them, the list of high-importance feature combinations is obtained by repeated offline training of the model, based on the comprehensive evaluation of the model effect (RMSE between the predicted value and the true value) and the importance of the features. In this embodiment, the algorithms used for the evaluation include but are not limited to traditional machine learning algorithms such as random forests and linear regression. The 10 groups of feature combinations with the highest evaluation importance (the actual number will be adjusted according to the actual situation) will be used as the list of high-importance feature combinations.

[0111] The combined feature construction method in the specific embodiment of the present invention is as follows:

[0112] 1. Pick any two existing parameters

[0113] 2. Combine the parameters into one parameter by string concatenation (e.g. the geographical location "suburbs" and the transaction volume "high" are combined into "suburbs_high")

[0114] 3. Evaluate the importance of combined features through algorithms that support feature importance evaluation, including but not limited to linear regression and RandomForest, which have fast training speeds. Combined features that exceed a certain threshold will be added to the original features as a supplement, otherwise they will be ignored.

[0115] 4. Return to step 1 and execute until the preset round.

[0116] In this embodiment, the feature supplement combination feature refers to the combination feature constructed and used by the method steps.

[0117] The cluster analysis module 3 analyzes the formatted data obtained by the preprocessing module 2 to obtain higher-level feature data, and obtains expert feature data by combining professional experience and banknote adding rules, and transmits it to the prediction algorithm module 4.

[0118] In the embodiment of the present invention, the ATM device has many unique data attributes. In terms of ATM transaction amount, according to the deposit amount, the ATM device can be divided into large deposits, relatively large deposits, relatively small deposits, and almost no deposits. Similarly, ATMs can also be classified from the perspective of withdrawal amount and deposit and withdrawal netting amount. In terms of ATM geographical location, ATM devices can also be divided into urban areas, suburbs, residential areas, commercial areas, etc. Classification by multiple dimensions will lead to a geometric increase in the combination of ATM categories, and too many ATM categories will make it difficult for expert experience to intervene.

[0119] The embodiment of the present invention adopts clustering technology to reduce the categories of ATM devices to an appropriate number while minimizing manual intervention, providing good conditions for the intervention of business experts. The additional cash adding rule information provided by business experts will greatly improve the modeling effect.

[0120] Figure 4 FIG. 3 is a structural block diagram of a cluster analysis module 3 of the ATM device banknote adding design auxiliary system according to an embodiment of the present invention. Figure 4 As shown, it includes:

[0121] The data clustering unit 31 performs clustering processing on the category data of the ATM to reduce the number of categories. In this embodiment, the clustering algorithms used by the data clustering unit 31 include but are not limited to common clustering algorithms such as K-Means, EM, and DBSCAN. In this embodiment, the K-Means algorithm is used to reduce the categories of the ATM devices to a number that business experts consider to be more appropriate, using the basic features of the ATM devices as input. In this embodiment of the present invention, the preset business expert rules are used to determine whether the number of categories obtained by clustering the ATM devices is at an appropriate level, otherwise the clustering analysis is performed again.

[0122] The business rule unit 32 will introduce the business experience of business experts based on the ATM categories of the clustering unit 31, and generate corresponding expert features for each category of ATM equipment by loading the pre-sorted expert rules, including but not limited to the priority of ATM equipment adding cash, the degree of influence by payday, the degree of influence by holidays, etc. The intervention of expert rules will provide the model with higher-level modeling information than the basic data features.

[0123] Specifically, a class of ATM machines obtained by performing a clustering algorithm in an embodiment of the present invention is specifically described as follows:

[0124] 1. If a certain type of ATM machines are all cash withdrawal machines next to a factory, then this type of machine should be greatly affected by payday. In the embodiment of the present invention, the factor affecting payday for this type of ATM machine will be set to a higher value;

[0125] 2. If a certain type of ATM machines has a large withdrawal volume on weekdays, but is located next to a software technology park, in an embodiment of the present invention, a higher value will be set for the extent to which such machines are affected by holidays based on transaction flows.

[0126] The prediction algorithm module 4 is responsible for the training and prediction of the intelligent model. The training of the intelligent model can be divided into a single algorithm model training unit 41, a preferred model unit 42 and a combined weight optimization unit 43. The prediction of the intelligent model is the responsibility of the plan suggestion generation unit 44. The input features of the prediction algorithm module 4 include the basic feature data of the ATM and the basic feature data of the personnel provided by the preprocessing module 2, and the high-level feature data and expert feature data provided by the cluster analysis module 3, while the prediction label is the design error of the banknote adding plan provided by the preprocessing module 2.

[0127] Figure 5FIG. 4 is a block diagram of the prediction algorithm module 4 in an embodiment of the present invention. Figure 5 As shown: In the context of machine learning, hyperparameters are parameters whose values ​​are set before starting the learning process, rather than parameter data obtained through training. Usually, hyperparameters need to be optimized to select a set of optimal hyperparameters for the learning machine to improve the performance and effect of learning.

[0128] The single algorithm model training unit 41 is responsible for the model training of a single algorithm. The prediction algorithm module 4 has built-in multiple candidate "algorithms" including but not limited to: SVM, KNN, RandomForest, GBDT, xgboost, LSTM, and each algorithm has its own unique one or more "hyperparameters". This constitutes an algorithm pool, and each group (algorithm, hyperparameter) combination is one of the "optional algorithms". For each "optional algorithm", the same feature input and prediction label will be used to train the model in conjunction with the preset hyperparameters. The single algorithm model training unit 41 will maximize the use of machine resources, and according to the situation of the host machine resources, perform model training and effect evaluation of multiple "optional algorithms" in parallel.

[0129] Each algorithm has one or more unique "hyperparameters". Taking RandomForest as an example, the hyperparameters include the number of trees and the depth of the trees. For a certain type of data, after multiple model trainings, we can find several sets of hyperparameters that usually perform better (for example, the number of trees = 100, the depth = 20). In the embodiments of the present invention, these hyperparameters with better performance are selected as preset parameters and used in the model training of the single algorithm here.

[0130] The preferred model unit 42 is responsible for selecting the preferred model from all the single algorithm models according to certain rules. The selection method of this patent is divided into two steps. The first step is to select the single algorithm model with the best evaluation effect of each "algorithm" according to the "algorithm". The second step is to select the top 4 models of the single algorithm models selected in the first step.

[0131] The combined weight optimization unit 43 is responsible for providing appropriate weight settings for the preferred model to improve the prediction effect of the combined model. The preferred model unit 42 provides four preferred models. In the model prediction, the preferred model provides four prediction results, namely x1, x2, x3, and x4, and the actual result is y. The combined weight optimization unit 43 uses x1, x2, x3, and x4 as features and y as a label to train a linear regression model and obtain the optimal weight values ​​w1, w2, w3, and w4. In the plan suggestion generation unit 44, the prediction value of the preferred model is weighted averaged using the optimal weight vector to obtain the final combined prediction result.

[0132] The plan suggestion generation unit 44 is responsible for generating the cash addition plan suggestion in the interaction with the user interaction module 1. The user interaction module 1 will provide the plan suggestion generation unit 44 with the ATM equipment number, the basic characteristic data of the personnel and the currently designed cash addition amount. The plan suggestion generation unit 44 will interact with the preprocessing module 2 and the cluster analysis module 3, obtain the ATM basic characteristic data, high-level characteristic data, and expert characteristic data according to the ATM equipment number, and finally call the model to predict the cash addition plan suggestion.

[0133] The self-update module 5 will trigger the retraining of the intelligent model according to certain rules to achieve the effect of automatic model update. The self-update module of the embodiment of the present invention is bottom-up. The self-update module first goes deep into each single algorithm model in the prediction algorithm module 4 to perform single algorithm model self-update, and then performs combined weight self-update on the overall combined model of the prediction algorithm module 4.

[0134] Figure 6 FIG. 5 is a flow chart showing the update strategy of the self-update module 5 in an embodiment of the present invention. Figure 6 As shown, it includes:

[0135] Single algorithm model evaluation S501: Evaluate whether the prediction effect of the single algorithm model is stable. If the last training time of the single algorithm model reaches the preset maximum value or the prediction result of the single algorithm model has abnormal deviations for 3 consecutive days (specifically, the error between the predicted value and the true value is higher than 150% of the average error in the training data), the model will be kicked out of the preferred model list and the single algorithm model self-update S502 will be triggered to retrain the single algorithm model, otherwise enter the combined model evaluation S503.

[0136] Single algorithm model self-update S502: obtain the latest data to retrain the single algorithm models of all non-optimal models, update the existing single algorithm models, and add them into the optimal model according to the evaluation effect of the single algorithm models.

[0137] Combined model self-update S503: If the single algorithm model self-update S502 is performed or the combined model prediction effect is lower than the threshold set by the preset rule, the self-update module will adjust the combined model weight parameters, that is, retrain the linear regression model.

[0138] The embodiment of the present invention provides an ATM equipment cash addition design auxiliary system and method. Based on a machine learning algorithm, data analysis, supervised learning, clustering and other technologies are comprehensively used to fit the difference between the ATM cash addition plan designed by the business personnel in history and the actual amount of consumption for different types of ATM equipment and different business personnel. In the process of designing the cash addition plan, auxiliary equipment management evaluation suggestions are given to the business personnel, guiding the business personnel to optimize the ATM cash addition plan, improving the accuracy of the cash addition plan, and directly improving the equipment management level of the business personnel.

[0139] At the same time, the present invention also provides a bank teller terminal adding money suggestion determining device, such as Figure 7 As shown, it includes:

[0140] Interaction module 701, used to obtain terminal device attribute data of the bank teller terminal to be loaded with banknotes, information of the banknote loading personnel and preset banknote loading plan data;

[0141] The data processing module 702 is used to determine the terminal device adding banknote characteristic data according to the terminal device attribute data and the banknote adding personnel information data;

[0142] The recommendation data generation module 703 is used to determine the terminal cash adding plan recommendation data based on the terminal equipment cash adding characteristic data, the preset cash adding plan data and the pre-trained cash adding recommendation model; wherein the cash adding recommendation model is a model established by machine learning based on the pre-acquired terminal characteristic data and cash adding personnel characteristic data.

[0143] In the embodiment of the present invention, the terminal equipment cash adding characteristic data includes: basic characteristic data of the terminal equipment, basic characteristic data of the cash adding personnel, clustering characteristic data and expert characteristic data; wherein,

[0144] The data processing module comprises:

[0145] A preprocessing unit, used for constructing features of the terminal device attribute data and the banknote adding personnel information data to determine the basic feature data of the terminal device and the basic feature data of the banknote adding personnel;

[0146] A clustering unit, used for clustering the basic characteristic data of the terminal device and the basic characteristic data of the cash adding personnel to generate cluster characteristic data;

[0147] The rule processing unit is used to determine the expert feature data according to the preset expert rules and the cluster feature data.

[0148] In an embodiment of the present invention, the device further comprises: a model training module, which is used to perform machine learning to establish a banknote adding suggestion model based on the previously acquired historical terminal device banknote adding feature data and historical banknote adding plan data; the model training module comprises:

[0149] A model training unit, used to train a candidate machine learning model of preset hyperparameters by using the previously acquired historical terminal device note adding feature data as input features and the historical note adding plan data as prediction labels;

[0150] A preferred model unit, used to determine a preferred machine learning model from the candidate machine learning models according to the model training result;

[0151] The weight optimization unit is used to determine the banknote adding suggestion model according to the preset weight setting and the determined preferred machine learning model.

[0152] In an embodiment of the present invention, the device further comprises:

[0153] An updating module is used to determine that if the training time of a single model in the preferred machine learning model is greater than a preset time threshold and / or the prediction results of a single algorithm model have abnormal deviations for more than a preset number of consecutive days, the single algorithm model will be removed from the preferred machine learning model.

[0154] For those skilled in the art, the implementation method of the device for determining banknote addition suggestion at a bank teller terminal provided by the present invention can be clearly known from the description of the above embodiments, which will not be described in detail here.

[0155] It should be noted that the method and device for determining bank teller terminal cash addition suggestions disclosed in the present invention can be used in the financial field or the artificial intelligence field, can be used for teller terminal cash addition in the financial field, and can also be used in any field other than the financial field. The application field of the method and device for determining bank teller terminal cash addition suggestions disclosed in the present invention is not limited.

[0156] This embodiment also provides an electronic device, which may be a desktop computer, a tablet computer, a mobile terminal, etc., but this embodiment is not limited thereto. In this embodiment, the electronic device may refer to the embodiments of the aforementioned method and device, the contents of which are incorporated herein, and the repeated parts are not repeated.

[0157] Figure 8 FIG. 6 is a schematic block diagram of a system structure of an electronic device 600 according to an embodiment of the present invention. Figure 8 As shown, the electronic device 600 may include a central processor 100 and a memory 140; the memory 140 is coupled to the central processor 100. It should be noted that this figure is exemplary; other types of structures may also be used to supplement or replace this structure to implement telecommunication functions or other functions.

[0158] In one embodiment, the bank teller terminal cash addition suggestion determination function may be integrated into the central processor 100. The central processor 100 may be configured to perform the following control:

[0159] Obtain terminal device attribute data of the bank teller terminal to be loaded with banknotes, information about the banknote loading personnel, and preset banknote loading plan data;

[0160] Determine the terminal device adding banknote characteristic data according to the terminal device attribute data and the banknote adding personnel information data;

[0161] The terminal adding banknotes plan suggestion data is determined according to the terminal equipment adding banknotes characteristic data, the preset adding banknotes plan data and the pre-trained adding banknotes suggestion model; wherein the adding banknotes suggestion model is a model established by machine learning based on the pre-acquired terminal characteristic data and the adding banknotes personnel characteristic data.

[0162] In another embodiment, the bank teller terminal cash addition suggestion determination device can be configured separately from the central processor 100. For example, the bank teller terminal cash addition suggestion determination processing device can be configured as a chip connected to the central processor 100, and the bank teller terminal cash addition suggestion determination processing function is realized through the control of the central processor.

[0163] like Figure 8 As shown, the electronic device 600 may further include: a communication module 110, an input unit 120, an audio processing unit 130, a display 160, and a power supply 170. It is worth noting that the electronic device 600 does not necessarily have to include Figure 8 In addition, the electronic device 600 may also include Figure 8 For components not shown, reference may be made to the prior art.

[0164] like Figure 8 As shown, the central processor 100 is sometimes also referred to as a controller or an operation control, and may include a microprocessor or other processor devices and / or logic devices. The central processor 100 receives inputs and controls the operations of various components of the electronic device 600.

[0165] The memory 140 may be, for example, one or more of a cache, a flash memory, a hard drive, a removable medium, a volatile memory, a non-volatile memory or other suitable devices. The above-mentioned information related to the failure may be stored, and a program for executing the relevant information may also be stored. The CPU 100 may execute the program stored in the memory 140 to implement information storage or processing.

[0166] The input unit 120 provides input to the CPU 100. The input unit 120 is, for example, a key or a touch input device. The power supply 170 is used to provide power to the electronic device 600. The display 160 is used to display display objects such as images and text. The display may be, for example, an LCD display, but is not limited thereto.

[0167] The memory 140 may be a solid-state memory, such as a read-only memory (ROM), a random access memory (RAM), a SIM card, etc. It may also be a memory that saves information even when the power is off, can be selectively erased, and is provided with more data, examples of which are sometimes referred to as EPROMs, etc. The memory 140 may also be some other type of device. The memory 140 includes a buffer memory 141 (sometimes referred to as a buffer). The memory 140 may include an application / function storage unit 142, which is used to store application programs and function programs or processes for executing the operation of the electronic device 600 through the central processor 100.

[0168] The memory 140 may also include a data storage unit 143 for storing data, such as contacts, digital data, pictures, sounds, and / or any other data used by the electronic device. The driver storage unit 144 of the memory 140 may include various drivers for communication functions of the electronic device and / or for executing other functions of the electronic device (such as messaging applications, address book applications, etc.).

[0169] The communication module 110 is a transmitter / receiver 110 that transmits and receives signals via an antenna 111. The communication module (transmitter / receiver) 110 is coupled to the central processor 100 to provide input signals and receive output signals, which may be the same as the case of a conventional mobile communication terminal.

[0170] Based on different communication technologies, multiple communication modules 110 may be provided in the same electronic device, such as a cellular network module, a Bluetooth module and / or a wireless LAN module. The communication module (transmitter / receiver) 110 is also coupled to a speaker 131 and a microphone 132 via an audio processor 130 to provide an audio output via the speaker 131 and receive an audio input from the microphone 132, thereby realizing a common telecommunication function. The audio processor 130 may include any suitable buffer, decoder, amplifier, etc. In addition, the audio processor 130 is also coupled to the central processor 100, so that the sound can be recorded on the local machine through the microphone 132, and the sound stored on the local machine can be played through the speaker 131.

[0171] An embodiment of the present invention further provides a computer-readable program, wherein when the program is executed in an electronic device, the program enables a computer to execute the bank teller terminal cash addition suggestion determination processing method as described in the above embodiment in the electronic device.

[0172] An embodiment of the present invention further provides a storage medium storing a computer-readable program, wherein the computer-readable program enables a computer to execute the bank teller terminal cash addition suggestion determination process described in the above embodiment in an electronic device.

[0173] The preferred embodiments of the present invention have been described above with reference to the accompanying drawings. Many features and advantages of these embodiments are clear from this detailed description, and therefore the appended claims are intended to cover all such features and advantages of these embodiments that fall within their true spirit and scope. In addition, since many modifications and changes are readily apparent to those skilled in the art, the embodiments of the present invention are not intended to be limited to the precise structure and operation illustrated and described, but all suitable modifications and equivalents that fall within their scope may be included.

[0174] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0175] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0176] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0177] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0178] The present invention uses specific embodiments to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for determining banknote addition suggestions at a bank teller terminal, characterized in that: The method includes: Obtain terminal device attribute data of the bank teller terminal to be loaded with banknotes, information about the banknote loading personnel, and preset banknote loading plan data; Determine the terminal device adding banknote characteristic data according to the terminal device attribute data and the banknote adding personnel information data; Determine the terminal adding banknote plan suggestion data according to the terminal equipment adding banknote characteristic data, the preset adding banknote plan data and the pre-trained adding banknote suggestion model; The terminal device attribute data includes: terminal transaction data, weather data of the location of the terminal device, date of adding banknotes, payday and repayment date data; The information of the cash adding personnel includes: identity information, working years and professional information of the cash adding personnel; The terminal equipment cash adding characteristic data includes: basic characteristic data of the terminal equipment, basic characteristic data of the cash adding personnel, cluster characteristic data and expert characteristic data; wherein, The method of determining the terminal device adding banknote characteristic data according to the terminal device attribute data and the banknote adding personnel information data includes: Performing feature construction on the terminal device attribute data and the cash adding personnel information data to determine the basic feature data of the terminal device and the basic feature data of the cash adding personnel; Performing clustering processing on the basic characteristic data of the terminal device and the basic characteristic data of the cash adding personnel to generate cluster characteristic data; Determine expert feature data according to preset expert rules and cluster feature data; the expert feature data at least includes: feature data of ATM device cash addition priority, feature data of the extent affected by payday, and feature data of the extent affected by holidays; The method further includes: establishing a banknote adding suggestion model by machine learning based on the historical banknote adding feature data of the terminal equipment and the historical banknote adding plan data acquired in advance; which includes: The previously acquired historical terminal device note adding feature data is used as input features, and the historical note adding plan data is used as prediction labels to train a candidate machine learning model of preset hyperparameters; Determine a preferred machine learning model from the candidate machine learning models according to the model training results; The banknote adding suggestion model is determined according to the preset weight setting and the determined preferred machine learning model.

2. The bank teller terminal cash addition suggestion determination method as claimed in claim 1, characterized in that: The method further comprises: If it is determined that the training time of a single algorithm model in the preferred machine learning model is greater than a preset time threshold and / or the prediction results of the single algorithm model have abnormal deviations for more than a preset number of consecutive days, the single algorithm model will be removed from the preferred machine learning model.

3. A bank teller terminal cash addition suggestion determination device, characterized in that: The device comprises: An interactive module, used to obtain terminal device attribute data of the bank teller terminal to be loaded with banknotes, information about the banknote loading personnel, and preset banknote loading plan data; A data processing module, used to determine the terminal device adding banknote characteristic data according to the terminal device attribute data and the banknote adding personnel information data; A suggestion data generation module is used to determine the terminal adding banknote plan suggestion data according to the terminal equipment adding banknote characteristic data, the preset adding banknote plan data and the pre-trained adding banknote suggestion model; wherein the adding banknote suggestion model is a model established by machine learning based on the pre-acquired terminal characteristic data and the adding banknote personnel characteristic data; The terminal device attribute data includes: terminal transaction data, weather data of the location of the terminal device, date of adding banknotes, payday and repayment date data; The information of the cash adding personnel includes: identity information, working years and professional information of the cash adding personnel; The terminal equipment cash adding characteristic data includes: basic characteristic data of the terminal equipment, basic characteristic data of the cash adding personnel, cluster characteristic data and expert characteristic data; wherein, The data processing module comprises: A preprocessing unit, used for constructing features of the terminal device attribute data and the banknote adding personnel information data to determine the basic feature data of the terminal device and the basic feature data of the banknote adding personnel; A clustering unit, used for clustering the basic characteristic data of the terminal device and the basic characteristic data of the cash adding personnel to generate cluster characteristic data; A rule processing unit, used to determine expert feature data according to preset expert rules and cluster feature data; the expert feature data at least includes: feature data of ATM device cash addition priority, feature data of the range affected by payday, and feature data of the range affected by holidays; The device further comprises: a model training module, which is used to perform machine learning to establish a banknote adding suggestion model based on the previously acquired historical terminal device banknote adding feature data and historical banknote adding plan data; the model training module comprises: A model training unit, used to train a candidate machine learning model of preset hyperparameters by using the previously acquired historical terminal device note adding feature data as input features and the historical note adding plan data as prediction labels; A preferred model unit, used to determine a preferred machine learning model from the candidate machine learning models according to the model training result; The weight optimization unit is used to determine the banknote adding suggestion model according to the preset weight setting and the determined preferred machine learning model.

4. The device for determining banknote addition suggestion at a bank teller terminal as claimed in claim 3, characterized in that: The device also includes: An updating module is used to determine that if the training time of a single algorithm model in the preferred machine learning model is greater than a preset time threshold and / or the prediction results of the single algorithm model have abnormal deviations for more than a preset number of consecutive days, the single algorithm model will be removed from the preferred machine learning model.

5. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 2 is implemented.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program for executing the method according to any one of claims 1 to 2.

Images