Cash management method and device for bank self-service equipment
By establishing a cash demand and inventory forecast model for bank self-service equipment and generating a clearing plan report in combination with the equipment type, the problems of inaccurate cash occupation and prediction in the existing technology are solved, and lower risks and costs are achieved.
Patent Information
- Application Number
- CN202210017219.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-01-07
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2042-01-07
AI Technical Summary
The cash clearing management method of existing bank self-service equipment relies on manual experience or simple calculations, resulting in an increase in cash occupation on the machine, making it impossible to accurately predict customer needs, and cannot update in time, resulting in high risks and costs.
By obtaining historical transaction data, customer composition data and external environment data, a cash demand prediction model and inventory prediction model are established, combined with equipment types, a clearing plan report is generated, and the model is retrained in real time is monitored and retrained to improve prediction accuracy.
Reduces cash occupation on board, reduces risks and costs, and improves the accuracy and timeliness of cash management.
Smart Images

Figure CN114202412B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of financial technology, and in particular to a cash management method and device for a bank self-service device. Background Art
[0002] As the burden on bank branches is reduced, the cash reserve management of self-service equipment has gradually shifted from branch staff to the centralized equipment management team. However, due to the limited staff of the centralized equipment management team and the large number of self-service equipment, it is easy to lead to extensive management of the business due to the small number of people and large number of machines.
[0003] In the prior art, the cash clearing management method of self-service equipment usually predicts cash demand based on manual experience thresholds, or predicts cash demand through simple calculation formulas, which can easily cause cash occupation on the machine, increase risks and costs, and do not distinguish between cash demands for different cash usage situations, resulting in the prediction results of some self-service equipment failing to meet actual customer needs. Cash demand is only predicted through historical transactions, which can easily lead to inaccurate prediction results. In addition, cash demand cannot be improved and updated in a timely manner.
[0004] It can be seen from this that the existing self-service cash clearing management method is likely to cause cash on the machine to be occupied and cannot reduce risks and costs. Summary of the invention
[0005] In view of this, an embodiment of the present invention provides a cash management method and device for a bank self-service device to achieve the purpose of reducing cash occupancy on the machine and reducing risks and costs.
[0006] To achieve the above objectives, the embodiments of the present invention provide the following technical solutions:
[0007] A first aspect of an embodiment of the present invention discloses a cash management method for a bank self-service device, the method comprising:
[0008] Acquire historical transaction data, historical customer composition data, and historical external environment data of the bank self-service device to be predicted within a first preset time range, wherein the historical transaction data includes historical deposit amounts and historical withdrawal amounts of the bank self-service device to be predicted;
[0009] Input the historical transaction data, historical customer composition data and historical external environment data into a pre-established cash demand forecasting model for forecasting, and output a cash demand forecast value, wherein the cash demand forecast value includes a forecast net difference, a maximum outflow forecast value at a time point and a maximum inflow forecast value at a time point;
[0010] Inputting the historical transaction data into a pre-established cash inventory forecasting model for forecasting, and outputting the forecasted cash inventory;
[0011] Determine the type of the bank self-service device to be predicted based on the historical deposit amount and the historical withdrawal amount, the type of the bank self-service device to be predicted includes a deposit device or a withdrawal device;
[0012] For different types of bank self-service devices to be predicted, a machine clearing plan report corresponding to the bank self-service devices to be predicted is determined based on the predicted cash inventory and the cash demand forecast value.
[0013] Optionally, also include:
[0014] Using a pre-established monitoring model to monitor the cash demand forecasting model in real time, and obtaining an average error of the cash demand forecasting model;
[0015] When the average error of the cash demand forecasting model meets a preset threshold condition, the cash demand forecasting model is retrained, and the threshold condition is that when the average error of the transaction volume of the bank self-service equipment that has completed the forecast for Y consecutive days within X days exceeds the threshold value Z yuan, a flag bit 1 is set, X and Y are positive integers, Z is a positive real number, and X is greater than or equal to Y;
[0016] The cash inventory prediction model is monitored in real time using a pre-established monitoring model to obtain a prediction ratio of the cash inventory prediction model, wherein the prediction ratio is a ratio of the actual cash inventory of the bank self-service device for which the prediction has been completed to the predicted cash inventory;
[0017] When the prediction ratio of the cash inventory prediction model is within a preset ratio condition, the cash inventory prediction model is retrained, and the ratio condition is the ratio of days in which the prediction ratio is within different preset ratio ranges.
[0018] Optionally, the using a pre-established monitoring model to monitor the cash demand forecasting model in real time to obtain an average error of the cash demand forecasting model includes:
[0019] Obtain the historical netting difference, the predicted netting difference and the number of the predicted netting difference of a specific number of bank self-service devices that have completed the forecast within the second set time range, and calculate the average error of the netting difference of the cash demand forecasting model;
[0020] Obtain the historical maximum outflow values at a certain time point, the maximum outflow forecast value at a certain time point, and the number of the maximum outflow forecast values at a certain time point of a specific number of bank self-service devices that have completed the forecast within the second set time range, and calculate the average error of the maximum outflow value at a certain time point of the cash demand forecast model;
[0021] The historical maximum point inflow values, the predicted maximum point inflow values and the number of the predicted maximum point inflow values of a specific number of the bank self-service devices that have completed the forecast within the second set time range are obtained, and the average error of the maximum point inflow value of the cash demand forecasting model is calculated.
[0022] Optionally, the using a pre-established monitoring model to monitor the cash inventory forecasting model in real time to obtain a forecast proportion of the cash inventory forecasting model includes:
[0023] The actual cash inventory and the predicted cash inventory of each bank self-service device for which the prediction has been completed within each set time range are obtained, and the prediction proportion of the cash inventory prediction model is calculated.
[0024] Optionally, the process of pre-establishing a cash demand forecasting model includes:
[0025] Obtaining sample data of each bank self-service device for which prediction has been completed, the sample data including historical cash demand data, historical customer composition data and historical external environment data, the historical external environment data including at least holiday data, weather data and derivative data;
[0026] Performing dimensionality reduction processing on each of the sample data to obtain each sample data after dimensionality reduction;
[0027] Input each of the sample data after dimensionality reduction into the machine learning algorithm XGBoost for prediction, and obtain the cash demand prediction value corresponding to each of the sample data after dimensionality reduction;
[0028] For each of the sample data after dimensionality reduction, obtain the cash demand forecast value and the cash demand actual value corresponding to the sample data after dimensionality reduction, and obtain the ratio results corresponding to each of the sample data after dimensionality reduction;
[0029] Based on the ratio results and penalty weight coefficient adjustment schemes respectively corresponding to each of the sample data after dimensionality reduction, it is determined whether the cash demand forecast value corresponding to each of the sample data after dimensionality reduction has been optimized;
[0030] If yes, confirm the optimization coefficient of the cash demand forecasting model, and update the machine learning algorithm XGBoost to obtain an updated machine learning algorithm XGBoost;
[0031] The cash demand forecasting model is generated based on the sample data after dimensionality reduction, the updated machine learning algorithm XGBoost and the penalty weight coefficient adjustment scheme.
[0032] A second aspect of an embodiment of the present invention discloses a cash management device for a bank self-service device, the device comprising:
[0033] An acquisition module, used to acquire historical transaction data, historical customer composition data and historical external environment data of the bank self-service device to be predicted within a first preset time range, wherein the historical transaction data includes historical deposit amounts and historical withdrawal amounts of the bank self-service device to be predicted;
[0034] A first prediction module is used to input the historical transaction data, historical customer composition data and historical external environment data into a pre-established cash demand prediction model for prediction, and output a cash demand prediction value, wherein the cash demand prediction value includes a predicted net difference, a maximum outflow prediction value at a time point and a maximum inflow prediction value at a time point;
[0035] A second prediction module is used to input the historical transaction data into a pre-established cash inventory prediction model for prediction, and output the predicted cash inventory;
[0036] A judgment module, used for judging the type of the bank self-service device to be predicted based on the historical deposit amount and the historical withdrawal amount, wherein the type of the bank self-service device to be predicted includes a deposit device or a withdrawal device;
[0037] The determination module is used to determine, for different types of bank self-service devices to be predicted, a machine clearing plan report corresponding to the bank self-service devices to be predicted based on the predicted cash inventory and the predicted cash demand value.
[0038] Optionally, also include:
[0039] A first monitoring module is used to monitor the cash demand forecasting model in real time using a pre-established monitoring model to obtain an average error of the cash demand forecasting model;
[0040] A first retraining module is used to retrain the cash demand forecasting model when the average error of the cash demand forecasting model meets a preset threshold condition, wherein the threshold condition is to set a flag bit 1 when the average error of the transaction volume of the bank self-service device that has completed the forecast for Y consecutive days within X days exceeds a threshold value of Z yuan, where X and Y are positive integers, Z is a positive real number, and X is greater than or equal to Y;
[0041] A second monitoring module is used to monitor the cash inventory prediction model in real time using a pre-established monitoring model to obtain a prediction ratio of the cash inventory prediction model, where the prediction ratio is a ratio of the actual cash inventory of the bank self-service device for which the prediction has been completed to the predicted cash inventory;
[0042] The second retraining module is used to retrain the cash inventory forecasting model when the prediction ratio of the cash inventory forecasting model is within a preset ratio condition, and the ratio condition is the ratio of days in which the prediction ratio is within different preset ratio ranges.
[0043] Optionally, the first monitoring module is specifically used to:
[0044] Obtain the historical net difference, the predicted net difference and the number of the predicted net difference of a specific number of bank self-service devices that have completed the forecast within the second set time range, and calculate the average error of the net difference of the cash demand forecasting model; obtain the historical point-in-time maximum outflow value, the point-in-time maximum outflow forecast value and the number of the point-in-time maximum outflow forecast values of a specific number of bank self-service devices that have completed the forecast within the second set time range, and calculate the average error of the point-in-time maximum outflow value of the cash demand forecasting model; obtain the historical point-in-time maximum inflow value, the point-in-time maximum inflow forecast value and the number of the point-in-time maximum inflow forecast values of a specific number of bank self-service devices that have completed the forecast within the second set time range, and calculate the average error of the point-in-time maximum inflow value of the cash demand forecasting model.
[0045] Optionally, the second monitoring module is specifically used to:
[0046] The actual cash inventory and the predicted cash inventory within each set time range of each bank self-service device for which the prediction has been completed are obtained, and the prediction proportion of the cash inventory prediction model is calculated.
[0047] Optionally, it also includes: pre-built modules;
[0048] The pre-built modules include:
[0049] An acquisition submodule is used to acquire sample data of each bank self-service device that has completed the forecast, wherein the sample data includes historical cash demand data, historical customer composition data and historical external environment data, wherein the historical external environment data includes at least holiday data, weather data and derivative data;
[0050] A dimensionality reduction processing submodule is used to perform dimensionality reduction processing on each of the sample data to obtain each sample data after dimensionality reduction;
[0051] A prediction submodule is used to input each of the sample data after dimensionality reduction into the machine learning algorithm XGBoost for prediction, so as to obtain a cash demand prediction value corresponding to each of the sample data after dimensionality reduction;
[0052] An obtaining submodule is used to obtain, for each of the sample data after dimensionality reduction, a cash demand forecast value and an actual cash demand value corresponding to the sample data after dimensionality reduction, and obtain ratio results corresponding to each of the sample data after dimensionality reduction;
[0053] A judgment submodule, used to judge whether the cash demand forecast value corresponding to each of the sample data after dimensionality reduction has been optimized based on the ratio results and penalty weight coefficient adjustment schemes respectively corresponding to each of the sample data after dimensionality reduction, and if so, execute an update submodule;
[0054] An updating submodule, used to confirm the optimization coefficient of the cash demand forecasting model and update the machine learning algorithm XGBoost to obtain an updated machine learning algorithm XGBoost;
[0055] A generation submodule is used to generate the cash demand forecasting model based on the sample data after dimensionality reduction, the updated machine learning algorithm XGBoost and the penalty weight coefficient adjustment scheme.
[0056] A cash management method and device for a bank self-service device provided based on the above-mentioned embodiment of the present invention, the method includes: obtaining historical transaction data, historical customer composition data and historical external environment data of the bank self-service device to be predicted within a first preset time range, the historical transaction data including the historical deposit amount and the historical withdrawal amount of the bank self-service device to be predicted; inputting the historical transaction data, the historical customer composition data and the historical external environment data into a pre-established cash demand forecasting model for forecasting, and outputting a cash demand forecast value, the cash demand forecast value including a forecast net difference, a point-in-time maximum outflow forecast value and a point-in-time maximum inflow forecast value; inputting the historical transaction data into a pre-established cash inventory forecasting model for forecasting, and outputting a forecast cash inventory; judging the type of the bank self-service device to be predicted based on the historical deposit amount and the historical withdrawal amount, the type of the bank self-service device to be predicted including a delivery device or a withdrawal device; for different types of bank self-service devices to be predicted, determining a machine clearing plan report corresponding to the bank self-service device to be predicted based on the forecast cash inventory and the cash demand forecast value. In this solution, a cash demand forecasting model and a cash inventory forecasting model are established in advance. Historical transaction data, historical customer composition data and historical external environment data are predicted in the cash demand forecasting model and the cash inventory forecasting model to obtain cash demand forecast values and predicted cash inventory respectively. The machine clearing plan report is determined according to the type of bank self-service equipment to be predicted, the predicted cash inventory and the cash demand forecast value, thereby reducing the cash occupancy on the machine and reducing risks and costs. BRIEF DESCRIPTION OF THE DRAWINGS
[0057] 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 embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.
[0058] Figure 1 A schematic diagram of a cash management method for a bank self-service device provided by an embodiment of the present invention;
[0059] Figure 2 A schematic diagram of a flow chart for pre-establishing a cash demand forecasting model provided by an embodiment of the present invention;
[0060] Figure 3 A schematic diagram of a prediction model provided by an embodiment of the present invention;
[0061] Figure 4 A schematic diagram of a cash management process of another bank self-service device provided by an embodiment of the present invention;
[0062] Figure 5 A schematic diagram of a process of pre-establishing a monitoring model provided by an embodiment of the present invention;
[0063] Figure 6 A schematic diagram of cash management of a bank self-service device provided by an embodiment of the present invention;
[0064] Figure 7 A schematic structural diagram of a cash management device for a bank self-service device provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0065] 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.
[0066] In this application, the terms "comprises", "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element.
[0067] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein.
[0068] In order to facilitate understanding of the technical solution of the present invention, the technical terms appearing in the present invention are explained:
[0069] On-board cash inventory: the cash balance in the cash box of the self-service device.
[0070] Adding cash to a self-service device: Put cash into the cash box of the self-service device to increase the cash balance in the device for customers to withdraw.
[0071] Cash return from self-service equipment: Take out cash from the cash box of the self-service equipment to reduce the cash balance in the equipment for customers to deposit.
[0072] Self-service equipment clearing: add money or return money to the self-service equipment.
[0073] Maximum outflow at a point in time: When customers make deposit or withdrawal transactions on self-service devices, deposit transactions are recorded as positive values, and withdrawal transactions are recorded as negative values. The amounts of each transaction are added together, and the maximum negative value reached within the time range is the maximum outflow at a point in time, which represents the maximum cash withdrawal demand of the customer.
[0074] Maximum point inflow: When customers make deposit and withdrawal transactions on self-service devices, deposit transactions are recorded as positive values, withdrawal transactions are recorded as negative values, and the amounts of each transaction are added up one by one. The maximum positive value reached within the time range is the maximum point inflow, which represents the maximum cash deposit demand of the customer.
[0075] Netting: When a customer makes a deposit or withdrawal transaction on a self-service device, the difference between the deposit amount and the withdrawal amount within a certain period of time is the netting.
[0076] Cash withdrawal equipment: Equipment whose cash deposit transaction amount is greater than the cash withdrawal transaction amount. It is often unable to continue to provide deposit transaction services because the cash box is full of cash.
[0077] Cash withdrawal equipment: Equipment where the cash withdrawal transaction amount is greater than the cash deposit transaction amount. Often, the equipment cannot provide withdrawal transaction services because there is no cash in the cash box.
[0078] Machine clearing plan: For all managed self-service devices, decide which devices to clear, whether to add or return cash to the clearing devices, and a detailed plan of the amount.
[0079] As can be seen from the background technology, the existing cash clearing management method of self-service equipment is likely to cause cash occupation on the machine and cannot reduce risks and costs.
[0080] Therefore, an embodiment of the present invention provides a cash management method and device for a bank self-service device. In this solution, a cash demand prediction model and a cash inventory prediction model are established in advance, and historical transaction data, historical customer composition data and historical external environment data are predicted in the cash demand prediction model and the cash inventory prediction model to obtain a cash demand prediction value and a predicted cash inventory, respectively. A machine clearing plan report is determined based on the type of bank self-service device to be predicted, the predicted cash inventory and the cash demand prediction value, thereby reducing the cash occupancy on the machine and reducing risks and costs.
[0081] like Figure 1 FIG. 1 is a flow chart of a cash management method for a bank self-service device provided by an embodiment of the present invention. The method mainly includes the following steps:
[0082] Step S101: Acquire historical transaction data, historical customer composition data and historical external environment data of the bank self-service device to be predicted within a first preset time range.
[0083] In step S101, optionally, the first preset time range may be 1 day, or 2 days, or 3 days, ...
[0084] The first preset time range is described below by way of example.
[0085] Assuming that the current time is December 20, 2021, and the predicted time is December 21, 2021, the first preset time range refers to the preset time range in the time range that is less than or equal to December 20, 2021. The preset time range can be 1 day or 2 days or 3 days or...
[0086] It should be noted that according to the cash usage of bank self-service equipment, bank self-service equipment can be divided into delivery equipment and withdrawal equipment. Therefore, the netting difference, the maximum outflow at a certain point in time and the maximum inflow at a certain point in time are set as the cash demand for different types of bank self-service equipment.
[0087] Optionally, in an embodiment of the present invention, historical transaction data includes but is not limited to historical netting of self-service devices, maximum outflows at historical points in time, and maximum inflows at historical points in time.
[0088] It is understandable that the historical transaction data includes but is not limited to the flow data of the bank self-service device to be predicted. For example, the historical transaction data includes the historical deposit amount and the historical withdrawal amount of the bank self-service device to be predicted.
[0089] The historical customer composition data includes, but is not limited to, customers who make deposits and withdrawals on holidays and customers who make deposits and withdrawals on weekdays.
[0090] Historical external environment data includes but is not limited to holiday data, weather data and derivative data.
[0091] Step S102: inputting historical transaction data, historical customer composition data and historical external environment data into a pre-established cash demand forecasting model for forecasting, and outputting a cash demand forecast value.
[0092] In step S102, the cash demand forecast value includes the forecast difference, the maximum forecast value of outflow at a certain point in time, and the maximum forecast value of inflow at a certain point in time.
[0093] Optionally, the cash demand prediction model predicts the cash demand of any bank self-service device to be predicted, that is, the cash demand prediction value output by the cash demand prediction model refers to the cash demand prediction value corresponding to the bank self-service device to be predicted.
[0094] Optionally, the cash demand forecasting model forecasts the cash demand of all the bank self-service devices to be forecasted at a bank branch, that is, the cash demand forecast value output by the cash demand forecasting model refers to the sum of the cash demand required by all the bank self-service devices to be forecasted at the bank branch.
[0095] In the specific implementation process of step S102, a cash demand forecasting model is pre-established, historical transaction data, historical customer composition data and historical external environment data are used as inputs of the established cash demand forecasting model, and the cash demand forecasting model is used to perform forecasting and output a cash demand forecast value.
[0096] Optionally, the actual net difference, the actual point-in-time maximum outflow value and the actual point-in-time maximum inflow value are respectively used as inputs of the established cash demand forecasting model, and the cash demand forecasting model is used for forecasting to output the predicted net difference, the predicted point-in-time maximum outflow value and the predicted point-in-time maximum inflow value.
[0097] Step S103: input the historical transaction data into a pre-established cash inventory forecasting model for forecasting, and output the forecasted cash inventory.
[0098] It should be noted that the execution principle and process of pre-establishing a cash inventory forecast model can be found in the execution principle and process of pre-establishing a cash demand forecast model, which will not be repeated here.
[0099] In the specific implementation process of step S103, a cash inventory forecasting model is pre-established, historical transaction data is used as input of the established cash inventory forecasting model, and forecasting is performed using the cash inventory forecasting model to output the forecasted cash inventory.
[0100] Optionally, the historical actual cash inventory is used as the input of the cash inventory forecasting model, and the cash inventory forecasting model is used to perform forecasting and output the forecasted cash inventory.
[0101] Step S104: Determine the type of the bank self-service device to be predicted based on the historical deposit amount and the historical withdrawal amount.
[0102] In step S104, the type of the bank self-service device to be predicted includes a delivery device or a return device.
[0103] In the process of specifically implementing step S104, it can be known from step S101 that the historical transaction data of the bank self-service device to be predicted within the first preset time range obtained includes the historical deposit amount and the historical withdrawal amount of the bank self-service device to be predicted, then the historical total deposit amount and the historical total withdrawal amount within the first preset time range can be determined, if the historical total deposit amount is greater than the historical total withdrawal amount, then the bank self-service device to be predicted is a recovery device, if the historical total withdrawal amount is greater than the historical total deposit amount, then the bank self-service device to be predicted is a delivery device.
[0104] Step S105: for different types of bank self-service devices to be predicted, determine a machine clearing plan report corresponding to the bank self-service devices to be predicted based on the predicted cash inventory and the predicted cash demand value.
[0105] In the process of specifically implementing step S105, when the predicted bank self-service device is a deposit device, it is determined whether the predicted bank self-service device needs to be cleared based on the predicted cash inventory and the predicted cash demand value. If necessary, the clearing plan report corresponding to the predicted bank self-service device is determined, and the bank self-service device to be predicted is added with cash or returned according to the clearing plan report; when the predicted bank self-service device is a return device, it is determined whether the predicted bank self-service device needs to be cleared based on the predicted cash inventory and the predicted cash demand value. If necessary, the clearing plan report corresponding to the predicted bank self-service device is determined, and the bank self-service device to be predicted is added with cash or returned according to the clearing plan report.
[0106] A cash management method for a bank self-service device provided based on an embodiment of the present invention comprises the following steps: obtaining historical transaction data, historical customer composition data and historical external environment data of a bank self-service device to be predicted within a first preset time range, wherein the historical transaction data includes historical deposit amounts and historical withdrawal amounts of the bank self-service device to be predicted; inputting the historical transaction data, historical customer composition data and historical external environment data into a pre-established cash demand prediction model for prediction, and outputting a cash demand prediction value, wherein the cash demand prediction value includes a predicted net difference, a maximum outflow prediction value at a time point and a maximum inflow prediction value at a time point; inputting the historical transaction data into a pre-established cash inventory prediction model for prediction, and outputting a predicted cash inventory; judging the type of the bank self-service device to be predicted based on the historical deposit amount and the historical withdrawal amount, wherein the type of the bank self-service device to be predicted includes a deposit device or a cash withdrawal device; and determining, for different types of bank self-service devices to be predicted, a machine clearing plan report corresponding to the bank self-service device to be predicted based on the predicted cash inventory and the predicted cash demand value. In this solution, a cash demand forecasting model and a cash inventory forecasting model are established in advance. Historical transaction data, historical customer composition data and historical external environment data are predicted in the cash demand forecasting model and the cash inventory forecasting model to obtain cash demand forecast values and predicted cash inventory respectively. The machine clearing plan report is determined according to the type of bank self-service equipment to be predicted, the predicted cash inventory and the cash demand forecast value, thereby reducing the cash occupancy on the machine and reducing risks and costs.
[0107] Based on the cash management method of a bank self-service device provided by the above embodiment of the present invention, step S102 is executed to input historical transaction data, historical customer composition data and historical external environment data into a pre-established cash demand forecasting model for forecasting, and output a cash demand forecast value, such as Figure 2 As shown, a flow chart of pre-establishing a cash demand forecasting model provided by an embodiment of the present invention mainly includes the following steps:
[0108] Step S201: Obtain sample data of each bank self-service device that has completed the prediction.
[0109] In step S201 , the sample data includes historical cash demand data, historical customer composition data, and historical external environment data.
[0110] Among them, historical external environment data at least includes holiday data, weather data and derivative data.
[0111] Optionally, the sample data also includes basic equipment information and historical transfer data, but is not limited to these.
[0112] Assume that the pre-point time is December 11, 2021, and each sample data is historical data of any time that is less than the current time.
[0113] Step S202: Perform dimensionality reduction processing on each sample data to obtain each sample data after dimensionality reduction.
[0114] It should be noted that by performing dimensionality reduction processing on each sample data, the model training time and required resources can be effectively reduced without affecting the machine learning algorithm.
[0115] Step S203: Input each of the reduced-dimensional sample data into the machine learning algorithm XGBoost for prediction, and obtain the cash demand prediction value corresponding to each of the reduced-dimensional sample data.
[0116] In the process of specifically implementing step S203, each sample data after dimensionality reduction is used as the input of the machine learning algorithm XGBoost, and each sample data after dimensionality reduction is predicted by the machine learning algorithm XGBoost to obtain the cash demand prediction value corresponding to each sample data after dimensionality reduction.
[0117] Step S204: for each sample data after dimensionality reduction, obtain the cash demand forecast value and the cash demand actual value corresponding to the sample data after dimensionality reduction, and obtain the ratio results corresponding to each sample data after dimensionality reduction.
[0118] In the process of specifically implementing step S204, for each sample data after dimensionality reduction, the cash demand forecast value and the actual cash demand value corresponding to the sample data after dimensionality reduction are obtained, and the actual cash demand value corresponding to each sample data after dimensionality reduction is divided by the cash demand forecast value to obtain the ratio results corresponding to each sample data after dimensionality reduction.
[0119] Step S205: Based on the ratio results and penalty weight coefficient adjustment schemes corresponding to each sample data after dimensionality reduction, determine whether the cash demand forecast value corresponding to each sample data after dimensionality reduction has been optimized. If so, execute step S206; if not, continue to execute step S205.
[0120] Optionally, after obtaining the ratio results corresponding to each sample data after dimensionality reduction, a threshold of 0.005 is set. When the ratio result is between [0.995 and 1.005], the cash demand forecast value is corrected, that is, the cash demand forecast value*ratio result.
[0121] That is to say, after obtaining the ratio results corresponding to each sample data after dimensionality reduction, the cash demand forecast value corresponding to each sample data after dimensionality reduction is optimized based on the ratio results corresponding to each sample data after dimensionality reduction and the penalty weight coefficient adjustment plan.
[0122] Step S206: confirm the optimization coefficient of the cash demand forecasting model, and update the machine learning algorithm XGBoost to obtain an updated machine learning algorithm XGBoost.
[0123] In the process of implementing step S206, it is determined that the penalty weight coefficient adjustment scheme has optimized the cash demand forecast value corresponding to each reduced-dimensional sample data, thereby determining the optimization coefficient of the cash demand forecast model, and updating the machine learning algorithm XGBoost to obtain an updated machine learning algorithm XGBoost.
[0124] Step S207: Generate a cash demand forecasting model based on the sample data after dimensionality reduction, the updated machine learning algorithm XGBoost and the penalty weight coefficient adjustment scheme.
[0125] In the specific implementation of step S207, a cash demand forecasting model is constructed using the sample data after dimensionality reduction, the updated machine learning algorithm XGBoost and the penalty weight coefficient adjustment scheme obtained above.
[0126] like Figure 3 As shown, it is a schematic diagram of a prediction model constructed according to an embodiment of the present invention.
[0127] exist Figure 3 In the process, historical transaction information, basic equipment information, historical transfer data, customer composition, holiday data, weather data and derivative data are firstly processed by dimension reduction to obtain the data after dimension reduction;
[0128] Next, the machine learning algorithm XGBoost is used to predict the data after dimensionality reduction to obtain the prediction results.
[0129] Next, the penalty weight coefficient adjustment scheme is used to determine whether the prediction results corresponding to each reduced-dimensional data have been optimized. If so, the optimization coefficient of the prediction model is confirmed, and the machine learning algorithm XGBoost is updated to obtain the updated machine learning algorithm XGBoost.
[0130] Finally, a prediction model is constructed using the reduced-dimensional data, the updated machine learning algorithm XGBoost, and the penalty weight coefficient adjustment scheme.
[0131] Preferably, the final prediction effect of XGBoost is shown in Table 1, where in the prediction of the transaction volume of self-service equipment, the model error of the maximum outflow at a time point reaches 21,000.
[0132] Table 1:
[0133]
[0134] Optionally, a cash inventory forecasting model may be constructed. The execution principle and process are the same as those of the above-mentioned construction of the forecasting model or the construction of the cash demand forecasting model, and will not be repeated here.
[0135] It should be noted that the cash inventory prediction model predicts the cash inventory on the self-service devices at the end of the day.
[0136] A cash management method for a bank self-service device provided in an embodiment of the present invention constructs a cash demand forecasting model by utilizing the sample data after dimensionality reduction, the updated machine learning algorithm XGBoost and the penalty weight coefficient adjustment scheme, thereby providing a guarantee for subsequent cash demand forecasting and achieving the purpose of reducing cash occupancy on board and reducing risks and costs.
[0137] Based on the cash management method of a bank self-service device provided by the above embodiment of the present invention, Figure 4 As shown, a flow chart of cash management of another bank self-service device provided by an embodiment of the present invention mainly includes the following steps:
[0138] Step S401: Acquire historical transaction data, historical customer composition data and historical external environment data of the bank self-service device to be predicted within a first preset time range.
[0139] Step S402: inputting historical transaction data, historical customer composition data and historical external environment data into a pre-established cash demand forecasting model for forecasting, and outputting a cash demand forecast value.
[0140] Step S403: input the historical transaction data into a pre-established cash inventory forecasting model for forecasting, and output the forecasted cash inventory.
[0141] Step S404: Determine the type of the bank self-service device to be predicted based on the historical deposit amount and the historical withdrawal amount.
[0142] Step S405: for different types of bank self-service devices to be predicted, determine a machine clearing plan report corresponding to the bank self-service devices to be predicted based on the predicted cash inventory and the predicted cash demand value.
[0143] It should be noted that the execution principle and process of the above steps S401 to S405 are the same as Figure 1 The execution principles and processes of steps S101 to S105 disclosed in are the same, which can be referred to and will not be repeated here.
[0144] Step S406: using the pre-established monitoring model to monitor the cash demand forecasting model in real time, and obtaining the average error of the cash demand forecasting model.
[0145] It should be noted that the evaluation dimensions of the forecasting model are the average model error and the forecast proportion. In other words, the evaluation dimension of the cash demand forecasting model is the average model error, and the evaluation dimension of the cash inventory forecasting model is the forecast proportion.
[0146] Optionally, executing step S406 is a process of using a pre-established monitoring model to monitor the cash demand forecasting model in real time to obtain an average error of the cash demand forecasting model, such as Figure 5 As shown, a schematic diagram of a process of pre-establishing a monitoring model provided by an embodiment of the present invention mainly includes the following steps:
[0147] Step S501: Obtain the historical netting, predicted netting and the number of predicted netting items of a specific number of bank self-service devices that have completed the forecast within the second set time range, and calculate the average error of the netting of the cash demand forecasting model.
[0148] In step S501, optionally, the second preset time range may be 1 day, or 2 days, or 3 days, ...
[0149] In an embodiment of the present invention,
[0150] Among them, x ij is the cash demand forecast value of the i-th bank self-service device on the j-th day / month / quarter, y ij is the actual cash demand value of the i-th bank self-service device that has completed the forecast on the j-th day / month / quarter, and n is the number of cash demand forecast values (the total number of forecasts).
[0151] Judging from the average error of the cash demand forecasting model, the smaller the average error of the cash demand forecasting model, the higher the accuracy of the cash demand forecasting model.
[0152] In the process of implementing step S501, the historical netting, predicted netting and the number of predicted netting items of a specific number of bank self-service devices that have completed the forecast within the second set time range are obtained. Based on the obtained historical netting, predicted netting and the number of predicted netting items, the average error calculation formula of the cash demand forecasting model is used to perform average error calculation to obtain the average error of the netting of the cash demand forecasting model.
[0153] For example, obtain the historical netting, predicted netting and the number of predicted netting items of 10 bank self-service devices that have completed the forecast within 2 days, and based on the obtained historical netting, predicted netting and the number of predicted netting items, use the average error calculation formula of the cash demand forecasting model to calculate the average error, and get the average error of the netting of the cash demand forecasting model.
[0154] Step S502: Obtain the historical maximum outflow values at a specific time point, the maximum outflow forecast value at a specific time point, and the number of maximum outflow forecast values at a specific time point of a specific number of bank self-service devices that have completed the forecast within the second set time range, and calculate the average error of the maximum outflow value at a specific time point of the cash demand forecast model.
[0155] In the process of specifically implementing step S502, the historical maximum outflow value at a certain point in time, the maximum outflow forecast value at a certain point in time and the number of maximum outflow forecast values at a certain point in time of a specific number of bank self-service devices that have completed the forecast within the second set time range are obtained. Based on the obtained historical maximum outflow value at a certain point in time, the maximum outflow forecast value at a certain point in time and the number of maximum outflow forecast values at a certain point in time, the average error calculation formula of the cash demand forecasting model is used to perform the average error calculation to obtain the average error of the maximum outflow value at a certain point in time of the cash demand forecasting model.
[0156] Step S503: Obtain the historical maximum point inflow values, the maximum point inflow forecast values, and the number of the maximum point inflow forecast values of a specific number of bank self-service devices that have completed the forecast within the second set time range, and calculate the average error of the maximum point inflow value of the cash demand forecasting model.
[0157] In the process of specifically implementing step S503, the historical maximum point inflow values, the maximum point inflow forecast values and the number of maximum point inflow forecast values of a specific number of bank self-service devices that have completed the forecast within the second set time range are obtained, and based on the obtained historical maximum point inflow values, the maximum point inflow forecast values and the number of maximum point inflow forecast values, the average error calculation formula of the cash demand forecasting model is used to perform average error calculation to obtain the average error of the maximum point inflow value of the cash demand forecasting model.
[0158] Step S407: When the average error of the cash demand forecasting model meets a preset threshold condition, the cash demand forecasting model is retrained.
[0159] In step S407, the threshold condition is to set the flag bit 1 when the average error of the transaction volume of the bank self-service device that has completed the prediction for Y consecutive days exceeds the threshold value Z yuan.
[0160] Wherein, X and Y are positive integers, Z is a positive real number, and X is greater than or equal to Y.
[0161] For example, the threshold condition is to set flag 1 when the average error of transaction volume of the bank self-service equipment that has completed the prediction for 8 consecutive days out of 10 days exceeds the threshold of 15,000 yuan.
[0162] In the specific implementation of step S407, after the average error of the cash demand forecasting model is obtained through the above calculation, a threshold condition is introduced to determine whether the average error of the cash demand forecasting model meets the preset threshold condition. If so, the cash demand forecasting model is retrained.
[0163] It can be understood that after obtaining the average error of the net difference of the cash demand forecasting model, or the average error of the maximum outflow at a time point, or the average error of the maximum inflow at a time point through the above calculation, the threshold condition is introduced to determine whether the average error of the net difference of the cash demand forecasting model, or the average error of the maximum outflow at a time point, or the average error of the maximum inflow at a time point meets the preset threshold condition. If so, the cash demand forecasting model is retrained.
[0164] Step S408: Use the pre-established monitoring model to monitor the cash inventory forecasting model in real time to obtain the forecast proportion of the cash inventory forecasting model.
[0165] In step S408, the predicted proportion is the ratio of the actual cash inventory of the bank's self-service equipment that has completed the prediction to the predicted cash inventory, that is:
[0166] Optionally, the process of executing step S408 to monitor the cash inventory forecast model in real time using a pre-established monitoring model to obtain the forecast proportion of the cash inventory forecast model includes:
[0167] The actual cash inventory and predicted cash inventory of each bank self-service device that has completed the forecast within each set time range are obtained, and the forecast proportion of the cash inventory forecast model is calculated.
[0168] Step S409: When the prediction ratio of the cash inventory prediction model is within a preset ratio condition, the cash inventory prediction model is retrained.
[0169] In step S409, the proportion condition is the proportion of days in which the predicted proportion is within different preset proportion ranges.
[0170] In the specific implementation of step S409, the predicted proportion of each bank self-service device that has completed the prediction is calculated for each day / month / quarter, and the proportion of days in different proportion ranges is calculated using each calculated predicted proportion.
[0171] It should be noted that the proportion of days in different proportion ranges is mainly based on the proportion range above which the proportion of days falls.
[0172] In the embodiment of the present invention, the proportion of days is mainly 0-50%, 50%-80%, 80%-100%, 100%-120%, 120%-150% and above 150%, among which, the three proportion ranges of 0-50%, 50%-80%, and 80%-100% represent that the predicted cash inventory is greater than the actual cash inventory, and at the ratio of 80%-120%, the more the better, that is, when the proportion of days is at the ratio of 80%-120%, there is no need to retrain the cash inventory prediction model.
[0173] It should be noted that the above process of calculating the average error of the cash demand forecasting model and the forecast proportion of the cash inventory forecasting model for each bank self-service device that has completed the forecast is the process of building a monitoring model report, and the monitoring model is built through the monitoring model report.
[0174] In the embodiment of the present invention, the cash demand forecasting model and the cash inventory forecasting model can be monitored in real time using the monitoring model report.
[0175] A cash management method for a bank self-service device provided by an embodiment of the present invention establishes a monitoring model in advance, and uses the established monitoring model to monitor the cash demand prediction model and the cash inventory prediction model. When the average error of the cash demand prediction model meets the preset threshold condition, the cash demand prediction model is retrained, and when the prediction ratio of the cash inventory prediction model is within the preset ratio condition, the cash inventory prediction model is retrained, thereby ensuring that the model effect is continuously accurate and available, thereby achieving the purpose of reducing the cash occupancy on the machine and reducing risks and costs.
[0176] In order to better understand the cash management method of a bank self-service device provided by the above embodiment of the present invention, Figure 6 As shown, it is a schematic diagram of cash management of a bank self-service device provided by an embodiment of the present invention.
[0177] exist Figure 6 In the process, the prediction model is first trained to generate a prediction model, and the prediction model is used to predict the cash demand of the bank's self-service equipment within the target time to obtain the corresponding cash demand prediction value, and a machine clearing plan is generated based on the obtained cash demand prediction value and the on-board cash inventory of the bank's self-service equipment within the current time.
[0178] Next, the monitoring model is used to monitor the prediction model in real time to obtain the error of the prediction model. When the error of the prediction model meets the threshold condition, the prediction model is retrained.
[0179] A cash management method for a bank self-service device provided in an embodiment of the present invention uses a generated prediction model to predict the cash demand of the bank self-service device within a target time, and generates a machine clearing plan based on the cash demand prediction value and the on-board cash inventory of the bank self-service device within the current time. At the same time, a monitoring model is used to monitor the prediction model in real time. When the error of the prediction model meets the threshold condition, the prediction model is retrained to ensure that the model effect is continuously accurate and available, thereby achieving the purpose of reducing the on-board cash occupancy and reducing risks and costs.
[0180] The above-mentioned embodiment of the present invention Figure 1 Corresponding to the cash management method of a bank self-service device shown in FIG, the embodiment of the present invention also provides a cash management device of a bank self-service device, such as Figure 7 As shown, the cash management device of the bank self-service equipment includes: an acquisition module 71, a first prediction module 72, a second prediction module 73, a judgment module 74 and a determination module 75.
[0181] The acquisition module 71 is used to acquire historical transaction data, historical customer composition data and historical external environment data of the bank self-service device to be predicted within a first preset time range.
[0182] The historical transaction data includes the historical deposit amount and historical withdrawal amount of the bank's self-service equipment to be predicted.
[0183] The first prediction module 72 is used to input historical transaction data, historical customer composition data and historical external environment data into a pre-established cash demand prediction model for prediction, and output a cash demand prediction value.
[0184] Among them, the cash demand forecast value includes the forecast difference, the maximum forecast value of outflow at a certain point in time, and the maximum forecast value of inflow at a certain point in time.
[0185] The second prediction module 73 is used to input the historical transaction data into a pre-established cash inventory prediction model for prediction, and output the predicted cash inventory.
[0186] The judgment module 74 is used to judge the type of the bank self-service device to be predicted based on the historical deposit amount and the historical withdrawal amount.
[0187] Among them, the types of bank self-service devices to be predicted include delivery devices or return devices.
[0188] The determination module 75 is used to determine the machine clearing plan report corresponding to the bank self-service equipment to be predicted based on the predicted cash inventory and the predicted cash demand value for different types of bank self-service equipment to be predicted.
[0189] It should be noted that the specific principles and execution processes of each module in the cash management device of the bank self-service equipment disclosed in the above-mentioned embodiment of the present invention are the same as the cash management method of the bank self-service equipment implemented by the above-mentioned present invention. Please refer to the corresponding parts of the cash management method of the bank self-service equipment disclosed in the above-mentioned embodiment of the present invention, and will not be repeated here.
[0190] A cash management method for a bank self-service device provided by an embodiment of the present invention pre-establishes a cash demand prediction model and a cash inventory prediction model, predicts historical transaction data, historical customer composition data and historical external environment data in the cash demand prediction model and the cash inventory prediction model, obtains a cash demand prediction value and a predicted cash inventory respectively, and determines a machine clearing plan report according to the type of bank self-service device to be predicted, the predicted cash inventory and the predicted cash demand value, thereby reducing cash occupancy on the machine and reducing risks and costs.
[0191] Optional, based on the above Figure 7 The cash management device of the bank self-service equipment shown is combined with Figure 7 The cash management device of the bank self-service equipment is further provided with a first monitoring module, a first retraining module, a second monitoring module and a second retraining module.
[0192] The first monitoring module is used to monitor the cash demand forecasting model in real time using a pre-established monitoring model to obtain an average error of the cash demand forecasting model.
[0193] The first retraining module is used to retrain the cash demand forecasting model when the average error of the cash demand forecasting model meets a preset threshold condition.
[0194] The threshold condition is that when the average error of the transaction volume of the bank self-service equipment that has completed the prediction for Y consecutive days exceeds the threshold value of Z yuan, the flag bit 1 is set, X and Y are positive integers, Z is a positive real number, and X is greater than or equal to Y.
[0195] The second monitoring module is used to monitor the cash inventory forecasting model in real time by using a pre-established monitoring model to obtain the forecast proportion of the cash inventory forecasting model.
[0196] Among them, the forecast ratio is the ratio of the actual cash inventory of the bank's self-service equipment that has completed the forecast to the forecast cash inventory.
[0197] The second retraining module is used to retrain the cash inventory forecasting model when the forecast proportion of the cash inventory forecasting model is within a preset proportion condition.
[0198] The proportion condition refers to the proportion of days in which the forecast proportion is within different preset proportion ranges.
[0199] Optional, based on the above Figure 7 In the cash management device of the bank self-service equipment shown, the first monitoring module is specifically used for:
[0200] Obtain the historical net difference, predicted net difference and the number of predicted net difference of a specific number of bank self-service devices that have completed the forecast within the second set time range, and calculate the average error of the net difference of the cash demand forecasting model; obtain the historical point-in-time maximum outflow value, point-in-time maximum predicted outflow value and the number of point-in-time maximum predicted outflow values of a specific number of bank self-service devices that have completed the forecast within the second set time range, and calculate the average error of the point-in-time maximum outflow value of the cash demand forecasting model; obtain the historical point-in-time maximum inflow value, point-in-time maximum predicted inflow value and the number of point-in-time maximum predicted inflow values of a specific number of bank self-service devices that have completed the forecast within the second set time range, and calculate the average error of the point-in-time maximum inflow value of the cash demand forecasting model.
[0201] Optional, based on the above Figure 7 The cash management device of the bank self-service equipment shown, the second monitoring module is specifically used for:
[0202] The actual cash inventory and predicted cash inventory of each bank self-service device that has completed the forecast within each set time range are obtained, and the forecast proportion of the cash inventory forecast model is calculated.
[0203] A cash management method for a bank self-service device provided by an embodiment of the present invention establishes a monitoring model in advance, and uses the established monitoring model to monitor the cash demand prediction model and the cash inventory prediction model. When the average error of the cash demand prediction model meets the preset threshold condition, the cash demand prediction model is retrained, and when the prediction ratio of the cash inventory prediction model is within the preset ratio condition, the cash inventory prediction model is retrained, thereby ensuring that the model effect is continuously accurate and available, thereby achieving the purpose of reducing the cash occupancy on the machine and reducing risks and costs.
[0204] Optional, based on the above Figure 7 The cash management device of the bank self-service equipment shown is combined with Figure 7 The cash management device of the bank's self-service equipment is further equipped with pre-built modules.
[0205] The pre-built module includes: an acquisition submodule, a dimensionality reduction processing submodule, a prediction submodule, a obtaining submodule, a judgment submodule, an updating submodule and a generating submodule.
[0206] The acquisition submodule is used to obtain sample data of each bank self-service device that has completed the prediction.
[0207] Among them, the sample data includes historical cash demand data, historical customer composition data and historical external environment data. The historical external environment data at least includes holiday data, weather data and derivative data.
[0208] The dimensionality reduction processing submodule is used to perform dimensionality reduction processing on each sample data to obtain each sample data after dimensionality reduction.
[0209] The prediction submodule is used to input each sample data after dimensionality reduction into the machine learning algorithm XGBoost for prediction, and obtain the cash demand prediction value corresponding to each sample data after dimensionality reduction.
[0210] A submodule is obtained, which is used to obtain, for each sample data after dimensionality reduction, a cash demand forecast value and an actual cash demand value corresponding to the sample data after dimensionality reduction, and obtain ratio results corresponding to each sample data after dimensionality reduction.
[0211] The judgment submodule is used to judge whether the cash demand forecast value corresponding to each reduced-dimensional sample data has been optimized based on the ratio results and penalty weight coefficient adjustment schemes corresponding to each reduced-dimensional sample data. If so, the update submodule is executed.
[0212] The update submodule is used to confirm the optimization coefficient of the cash demand forecasting model and update the machine learning algorithm XGBoost to obtain an updated machine learning algorithm XGBoost.
[0213] The generation submodule is used to generate a cash demand forecasting model based on the sample data after dimensionality reduction, the updated machine learning algorithm XGBoost and the penalty weight coefficient adjustment scheme.
[0214] A cash management method for a bank self-service device provided in an embodiment of the present invention constructs a cash demand forecasting model by utilizing the sample data after dimensionality reduction, the updated machine learning algorithm XGBoost and the penalty weight coefficient adjustment scheme, thereby providing a guarantee for subsequent cash demand forecasting and achieving the purpose of reducing cash occupancy on board and reducing risks and costs.
[0215] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can refer to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiment. The system and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without creative work.
[0216] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.
[0217] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A cash management method for a bank self-service device, characterized in that: The method comprises: Acquire historical transaction data, historical customer composition data, and historical external environment data of the bank self-service device to be predicted within a first preset time range, wherein the historical transaction data includes historical deposit amounts and historical withdrawal amounts of the bank self-service device to be predicted; Input the historical transaction data, historical customer composition data and historical external environment data into a pre-established cash demand forecasting model for forecasting, and output a cash demand forecast value, wherein the cash demand forecast value includes a forecast net difference, a maximum outflow forecast value at a time point and a maximum inflow forecast value at a time point; Inputting the historical transaction data into a pre-established cash inventory forecasting model for forecasting, and outputting the forecasted cash inventory; Determine the type of the bank self-service device to be predicted based on the historical deposit amount and the historical withdrawal amount, the type of the bank self-service device to be predicted includes a deposit device or a withdrawal device; For different types of bank self-service devices to be predicted, a machine clearing plan report corresponding to the bank self-service devices to be predicted is determined based on the predicted cash inventory and the cash demand forecast value.
2. The method according to claim 1, characterized in that Also includes: Using a pre-established monitoring model to monitor the cash demand forecasting model in real time, and obtaining an average error of the cash demand forecasting model; When the average error of the cash demand forecasting model meets a preset threshold condition, the cash demand forecasting model is retrained, and the threshold condition is that when the average error of the transaction volume of the bank self-service equipment that has completed the forecast for Y consecutive days within X days exceeds the threshold value Z yuan, a flag bit 1 is set, X and Y are positive integers, Z is a positive real number, and X is greater than or equal to Y; The cash inventory prediction model is monitored in real time using a pre-established monitoring model to obtain a prediction ratio of the cash inventory prediction model, wherein the prediction ratio is a ratio of the actual cash inventory of the bank self-service device for which the prediction has been completed to the predicted cash inventory; When the prediction ratio of the cash inventory prediction model is within a preset ratio condition, the cash inventory prediction model is retrained, and the ratio condition is the ratio of days in which the prediction ratio is within different preset ratio ranges.
3. The method according to claim 2, characterized in that The using of a pre-established monitoring model to monitor the cash demand forecasting model in real time to obtain an average error of the cash demand forecasting model includes: Obtain the historical netting difference, the predicted netting difference and the number of the predicted netting difference of a specific number of bank self-service devices that have completed the forecast within the second set time range, and calculate the average error of the netting difference of the cash demand forecasting model; Obtain the historical maximum outflow values at a certain time point, the maximum outflow forecast value at a certain time point, and the number of the maximum outflow forecast values at a certain time point of a specific number of bank self-service devices that have completed the forecast within the second set time range, and calculate the average error of the maximum outflow value at a certain time point of the cash demand forecast model; The historical maximum point inflow values, the predicted maximum point inflow values and the number of the predicted maximum point inflow values of a specific number of the bank self-service devices that have completed the forecast within the second set time range are obtained, and the average error of the maximum point inflow value of the cash demand forecasting model is calculated.
4. The method according to claim 2, characterized in that: The method of using a pre-established monitoring model to monitor the cash inventory forecasting model in real time to obtain a forecast proportion of the cash inventory forecasting model includes: The actual cash inventory and the predicted cash inventory of each bank self-service device for which the prediction has been completed within each set time range are obtained, and the prediction proportion of the cash inventory prediction model is calculated.
5. The method according to claim 1, characterized in that The process of pre-establishing a cash demand forecasting model includes: Obtaining sample data of each bank self-service device that has completed the forecast, the sample data includes historical cash demand data, historical customer composition data and historical external environment data, the historical external environment data at least includes holiday data, weather data and derivative data; Performing dimensionality reduction processing on each of the sample data to obtain each sample data after dimensionality reduction; Input each of the sample data after dimensionality reduction into the machine learning algorithm XGBoost for prediction, and obtain the cash demand prediction value corresponding to each of the sample data after dimensionality reduction; For each of the sample data after dimensionality reduction, obtain the cash demand forecast value and the cash demand actual value corresponding to the sample data after dimensionality reduction, and obtain the ratio results corresponding to each of the sample data after dimensionality reduction; Based on the ratio results and penalty weight coefficient adjustment schemes respectively corresponding to each of the sample data after dimensionality reduction, it is determined whether the cash demand forecast value corresponding to each of the sample data after dimensionality reduction has been optimized; If yes, confirm the optimization coefficient of the cash demand forecasting model, and update the machine learning algorithm XGBoost to obtain an updated machine learning algorithm XGBoost; The cash demand forecasting model is generated based on the sample data after dimensionality reduction, the updated machine learning algorithm XGBoost and the penalty weight coefficient adjustment scheme.
6. A cash management device for a bank self-service device, characterized in that: The device comprises: An acquisition module, used to acquire historical transaction data, historical customer composition data and historical external environment data of the bank self-service device to be predicted within a first preset time range, wherein the historical transaction data includes historical deposit amounts and historical withdrawal amounts of the bank self-service device to be predicted; A first prediction module is used to input the historical transaction data, historical customer composition data and historical external environment data into a pre-established cash demand prediction model for prediction, and output a cash demand prediction value, wherein the cash demand prediction value includes a predicted net difference, a maximum outflow prediction value at a time point and a maximum inflow prediction value at a time point; A second prediction module is used to input the historical transaction data into a pre-established cash inventory prediction model for prediction, and output the predicted cash inventory; A judgment module, used for judging the type of the bank self-service device to be predicted based on the historical deposit amount and the historical withdrawal amount, wherein the type of the bank self-service device to be predicted includes a deposit device or a withdrawal device; The determination module is used to determine, for different types of bank self-service devices to be predicted, a machine clearing plan report corresponding to the bank self-service devices to be predicted based on the predicted cash inventory and the predicted cash demand value.
7. The device according to claim 6, characterized in that Also includes: A first monitoring module is used to monitor the cash demand forecasting model in real time using a pre-established monitoring model to obtain an average error of the cash demand forecasting model; A first retraining module is used to retrain the cash demand forecasting model when the average error of the cash demand forecasting model meets a preset threshold condition, wherein the threshold condition is to set a flag bit 1 when the average error of the transaction volume of the bank self-service device that has completed the forecast for Y consecutive days within X days exceeds a threshold value of Z yuan, where X and Y are positive integers, Z is a positive real number, and X is greater than or equal to Y; A second monitoring module is used to monitor the cash inventory prediction model in real time using a pre-established monitoring model to obtain a prediction ratio of the cash inventory prediction model, where the prediction ratio is a ratio of the actual cash inventory of the bank self-service device for which the prediction has been completed to the predicted cash inventory; The second retraining module is used to retrain the cash inventory forecasting model when the prediction ratio of the cash inventory forecasting model is within a preset ratio condition, and the ratio condition is the ratio of days in which the prediction ratio is within different preset ratio ranges.
8. The device according to claim 7, characterized in that The first monitoring module is specifically used for: Obtain the historical net difference, the predicted net difference and the number of the predicted net difference of a specific number of bank self-service devices that have completed the forecast within the second set time range, and calculate the average error of the net difference of the cash demand forecasting model; obtain the historical point-in-time maximum outflow value, the point-in-time maximum outflow forecast value and the number of the point-in-time maximum outflow forecast values of a specific number of bank self-service devices that have completed the forecast within the second set time range, and calculate the average error of the point-in-time maximum outflow value of the cash demand forecasting model; obtain the historical point-in-time maximum inflow value, the point-in-time maximum inflow forecast value and the number of the point-in-time maximum inflow forecast values of a specific number of bank self-service devices that have completed the forecast within the second set time range, and calculate the average error of the point-in-time maximum inflow value of the cash demand forecasting model.
9. The device according to claim 7, characterized in that The second monitoring module is specifically used for: The actual cash inventory and the predicted cash inventory within each set time range of each bank self-service device for which the prediction has been completed are obtained, and the prediction proportion of the cash inventory prediction model is calculated.
10. The device according to claim 6, characterized in that Also includes: Pre-built modules; The pre-built modules include: An acquisition submodule is used to acquire sample data of each bank self-service device that has completed the forecast, wherein the sample data includes historical cash demand data, historical customer composition data and historical external environment data, wherein the historical external environment data includes at least holiday data, weather data and derivative data; A dimensionality reduction processing submodule is used to perform dimensionality reduction processing on each of the sample data to obtain each sample data after dimensionality reduction; A prediction submodule is used to input each of the sample data after dimensionality reduction into the machine learning algorithm XGBoost for prediction, so as to obtain a cash demand prediction value corresponding to each of the sample data after dimensionality reduction; An obtaining submodule is used to obtain, for each of the sample data after dimensionality reduction, a cash demand forecast value and an actual cash demand value corresponding to the sample data after dimensionality reduction, and obtain ratio results corresponding to each of the sample data after dimensionality reduction; A judgment submodule, used to judge whether the cash demand forecast value corresponding to each of the sample data after dimensionality reduction has been optimized based on the ratio results and penalty weight coefficient adjustment schemes respectively corresponding to each of the sample data after dimensionality reduction, and if so, execute an update submodule; An updating submodule, used to confirm the optimization coefficient of the cash demand forecasting model and update the machine learning algorithm XGBoost to obtain an updated machine learning algorithm XGBoost; A generation submodule is used to generate the cash demand forecasting model based on the sample data after dimensionality reduction, the updated machine learning algorithm XGBoost and the penalty weight coefficient adjustment scheme.
