CN112785063B - Transaction amount prediction system based on transaction amount prediction model - Google Patents
Transaction amount prediction system based on transaction amount prediction model Download PDFInfo
- Publication number
- CN112785063B CN112785063B CN202110105443.7A CN202110105443A CN112785063B CN 112785063 B CN112785063 B CN 112785063B CN 202110105443 A CN202110105443 A CN 202110105443A CN 112785063 B CN112785063 B CN 112785063B
- Authority
- CN
- China
- Prior art keywords
- transaction
- transaction amount
- weight
- data
- date
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/04—Trading; Exchange, e.g. stocks, commodities, derivatives or currency exchange
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Finance (AREA)
- Accounting & Taxation (AREA)
- Marketing (AREA)
- Development Economics (AREA)
- Human Resources & Organizations (AREA)
- General Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Technology Law (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a transaction amount prediction system based on a transaction amount prediction model, which relates to the field of data mining and comprises the following components: the transaction data acquisition module is used for acquiring historical business transaction amount data; the first transaction processing module is used for processing the historical business transaction amount data to obtain a first total transaction amount and an all-day transaction amount which correspond to the transaction date and are in a preset period as a training set; the prediction model training module is used for training according to the training set to obtain a transaction amount prediction model which takes the transaction date and the first total transaction amount as input and takes the total transaction amount as output; and the transaction amount prediction module is used for inputting the predicted date and the corresponding total transaction amount on the same day into the transaction amount prediction model to obtain the predicted total transaction amount on the same day, and processing the predicted total transaction amount according to the historical transaction amount data of a preset period before the predicted date and the predicted total transaction amount to obtain a transaction amount prediction result of the predicted date. The method has the beneficial effects that the accuracy of the business transaction amount prediction result is effectively improved.
Description
Technical Field
The invention relates to the field of data mining time sequences, in particular to a transaction amount prediction system based on a transaction amount prediction model.
Background
The t+0 settlement transaction service requires an enterprise's own fund pool for the payment made by the transaction service of each agent. Too small a setting of the self-contained fund pool can affect the development of the business, and too large a setting can result in wastage of the fund cost, so that a premature prediction of the transaction amount becomes necessary. The existing statistical methods such as moving average and exponential smoothing are too rough, and other methods such as seasonal decomposition and ARIMA require consistency of data time on the one hand and become less flexible on the other hand when considering the influence of related factors. Because the T+0 service has operational flexibility, the algorithm can not completely meet the actual needs under the condition that the weekend service is just started and the holiday service is closed.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a transaction amount prediction system based on a transaction amount prediction model, which comprises the following steps:
the transaction data acquisition module is used for acquiring historical service transaction amount data, wherein the historical service transaction amount data comprises a transaction date, a transaction period and historical service transaction amounts corresponding to the transaction date and the transaction period;
the first transaction processing module is connected with the transaction data acquisition module and is used for processing the historical business transaction amount data to obtain a first total transaction amount and an all-day transaction amount which correspond to the transaction date and are in a preset period as a training set;
the prediction model training module is connected with the first transaction processing module and is used for training according to the training set to obtain a transaction amount prediction model taking the transaction date and the first total transaction amount as input and the total transaction amount as output;
the transaction amount prediction module is respectively connected with the transaction data acquisition module and the prediction model training module and is used for inputting a predicted date and a corresponding total transaction amount on the same day into the transaction amount prediction model to obtain a predicted total-day transaction amount, and processing the historical transaction amount data in a preset period before the predicted date and the predicted total-day transaction amount to obtain a transaction amount prediction result of the predicted date.
Preferably, the first transaction processing module includes:
a first processing unit, configured to calculate a first average value of the first aggregate transaction amounts corresponding to the same transaction date in the same period, and add the first aggregate transaction amount greater than the first average value to a first data set, and add the first aggregate transaction amount not greater than the first average value to a second data set;
a second processing unit, connected to the first processing unit, for calculating a first duty ratio of the number of the first total amount of transactions in the first data set to the number of all the first total amount of transactions, and a second duty ratio of the number of the first total amount of transactions in the second data set to the number of all the first total amount of transactions, respectively;
the third processing unit is connected with the second processing unit and is used for storing the corresponding relation between the pre-configured duty ratio and the corresponding weight, matching the corresponding weight according to the first duty ratio to serve as a first weight, and matching the corresponding weight according to the second duty ratio to serve as a second weight;
and the fourth processing unit is respectively connected with the first processing unit and the third processing unit and is used for calculating a second average value of each first total transaction amount in the first data set and a third average value of each first total transaction amount in the second data set, and adding the third average value and the second weight into the training set as the first total transaction amount of a preset period corresponding to the transaction date according to the second average value and the first weight.
Preferably, the sum of the first weight and the second weight is 1.
Preferably, the first transaction processing module includes:
a fifth processing unit, configured to calculate a fourth average value of the all-day transaction amounts corresponding to the same transaction date in the same period, and add the all-day transaction amounts greater than the fourth average value to a third data set, and add the all-day transaction amounts not greater than the fourth average value to a fourth data set;
a sixth processing unit, connected to the fifth processing unit, configured to respectively calculate a third duty ratio of the number of all-day transaction amounts in the third data set to the number of all-day transaction amounts, and a fourth duty ratio of the number of all-day transaction amounts in the fourth data set to the number of all-day transaction amounts;
a seventh processing unit, connected to the sixth processing unit, configured to save a correspondence between a preconfigured duty ratio and a corresponding weight, and match the corresponding weight according to the third duty ratio as a third weight, and match the corresponding weight according to the fourth duty ratio as a fourth weight;
and an eighth processing unit, connected to the fifth processing unit and the seventh processing unit, respectively, configured to calculate a fifth average value of each of the all-day transaction amounts in the third data set and a sixth average value of each of the all-day transaction amounts in the fourth data set, and perform weighted summation according to the fifth average value and the third weight, and the sixth average value and the fourth weight, to be used as the all-day transaction amount corresponding to the transaction date to be added into the training set.
Preferably, the sum of the third weight and the fourth weight is 1.
Preferably, the system further comprises a second transaction processing module connected with the transaction data acquisition module and used for processing the historical service transaction amount data to obtain a first trend curve of the historical service transaction amount changing along with the transaction period under the transaction date;
the transaction amount prediction module includes:
a similarity calculating unit, configured to process a second trend curve of the historical transaction amount of the preset period before the predicted date according to the transaction period, and calculate a total similarity between the second trend curve and the first trend curve in the same period;
and the prediction unit is connected with the similarity calculation unit and is used for processing the total daily transaction amount of the transaction date corresponding to the total similarity and the predicted total daily transaction amount to obtain a service transaction amount prediction result of the predicted date.
Preferably, the prediction unit includes:
the storage subunit is used for storing at least one preset similarity interval, each similarity interval corresponds to a group of weight values, and the weight values comprise at least one fifth weight and one sixth weight;
and the calculation subunit is connected with the storage subunit and is used for obtaining the corresponding weight value according to the similarity matching, and carrying out weighted summation calculation according to the all-day transaction amount and the fifth weight and the predicted all-day transaction amount and the sixth weight to obtain a service transaction amount prediction result of the predicted date.
Preferably, the preset period includes five working days before the predicted date;
the similarity calculation unit includes:
a curve generation subunit, configured to generate the second trend curve of the historical transaction amount according to the transaction period for each of the five working days;
the first calculation subunit is connected with the curve generation subunit and is used for respectively calculating the sub-similarity of the second trend curve and the first trend curve in the same period, and arranging the sub-similarity according to the sequence from big to small to form a similarity queue;
the second calculation subunit is connected with the first calculation subunit and is used for storing five pre-configured weight queues which are formed by arranging the weight queues in a sequence from big to small, and carrying out corresponding sequencing position weighted summation according to the similarity queues and the weight queues to obtain the total similarity.
Preferably, the sum of the weights in the weight queue is 1.
Preferably, the system further comprises a business abnormality detection module which is respectively connected with the transaction data acquisition module and the first transaction processing module, wherein the business abnormality detection module comprises,
an initializing unit for initializing the historical business transaction amount data;
the training unit is connected with the initializing unit and learns the statistical rule of the historical business transaction amount data to obtain a rule set with an upper limit index and a lower limit index;
the abnormality detection unit is connected with the training unit and used for judging whether the historical service data amount is in the range of the online index and the offline index of the rule set; and
if so, the historical service transaction amount data is normal, otherwise, the historical service transaction amount data is abnormal;
and the first transaction processing module processes the normal historical business transaction limit data to obtain the training set.
The technical scheme has the following advantages or beneficial effects: and on the basis of adopting the total transaction amount of the day of the preset period to predict the total transaction amount, the historical business transaction amount data before the preset date is considered to predict the total transaction amount, so that the accuracy of the finally obtained business transaction amount prediction result is effectively improved.
Drawings
FIG. 1 is a schematic diagram of a transaction amount prediction system based on a transaction amount prediction model according to a preferred embodiment of the present invention.
Detailed Description
The invention will now be described in detail with reference to the drawings and specific examples. The present invention is not limited to the embodiment, and other embodiments may fall within the scope of the present invention as long as they conform to the gist of the present invention.
In accordance with the foregoing problems with the prior art, the present invention provides a transaction amount prediction system based on a transaction amount prediction model, as shown in fig. 1, comprising:
the transaction data acquisition module 1 is used for acquiring historical service transaction amount data, wherein the historical service transaction amount data comprises a transaction date, a transaction period and historical service transaction amounts corresponding to the transaction date and the transaction period;
the first transaction processing module 2 is connected with the transaction data acquisition module 1 and is used for processing the historical business transaction amount data to obtain a first total transaction amount and an all-day transaction amount of a preset period corresponding to a transaction date as a training set;
the prediction model training module 3 is connected with the first transaction processing module 2 and is used for training according to the training set to obtain a transaction amount prediction model which takes the transaction date and the first total transaction amount as input and takes the total transaction amount as output;
the transaction amount predicting module 4 is respectively connected with the transaction data acquiring module 1 and the predicting model training module 3, and is used for inputting the predicted date and the corresponding total transaction amount on the same day into the transaction amount predicting model to obtain the predicted total-day transaction amount, and processing the predicted total-day transaction amount according to the historical transaction amount data of a preset period before the predicted date and the predicted total-day transaction amount to obtain the service transaction amount predicting result of the predicted date.
Specifically, in this embodiment, the present application can implement prediction of the total daily transaction amount of the T0 settlement service, and the corresponding prediction time point is preferably between 14:00 and 14:30 on the same day, and after 13 hours of data generation; the above parameters are only one embodiment of the present invention, and specific values of the parameters can be set according to actual situations, so the present invention is not limited thereto.
Further, the preset period may be an afternoon period, and the transaction period and the corresponding historical transaction amount included in the preset period by the transaction data obtaining module 1 may be extracted by the historical transaction amount data obtained by the transaction data obtaining module 1, so as to obtain the sum of the historical transaction amounts of the preset period corresponding to the transaction date, that is, the first total transaction amount, and meanwhile, the total day transaction amount corresponding to the transaction date may be obtained by processing the historical transaction amounts of all the transaction periods in one day according to the transaction date. And then training by taking the transaction date and the corresponding first total transaction amount as input and taking the total transaction amount corresponding to the transaction date as output to obtain a transaction amount prediction model. After acquiring the total transaction amount of the predicted date on the current day corresponding to the preset period, the transaction amount prediction model can be used for preliminarily predicting to obtain a predicted total daily transaction amount of the predicted date, and because the predicted date may be in holidays or other service peak periods or may be in service valley periods, the prediction of the total daily transaction amount by independently adopting the transaction amount prediction model may not be accurate, and because the service peak or valley periods usually have tendency, i.e. the service is gradually increased or gradually decreased, no mutation occurs, in the embodiment, the variation trend of the total daily transaction amount can be acquired by combining the historical service transaction amount data of the preset period before the predicted date, thereby improving the accuracy of the service transaction amount prediction result.
In a preferred embodiment of the present invention, the first transaction processing module 2 comprises:
a first processing unit 21, configured to calculate a first average value of first aggregate transactions corresponding to the same transaction date in the same period, and add the first aggregate transaction amount greater than the first average value to a first data set, and add the first aggregate transaction amount not greater than the first average value to a second data set;
a second processing unit 22 connected to the first processing unit 21 for calculating a first duty ratio of the number of the first aggregate intersections in the first data set to the number of all the first aggregate intersections, and a second duty ratio of the number of the first aggregate intersections in the second data set to the number of all the first aggregate intersections, respectively;
a third processing unit 23, connected to the second processing unit 22, configured to store a correspondence between a pre-configured duty ratio and a corresponding weight, and match the corresponding weight according to the first duty ratio as a first weight, and match the corresponding weight according to the second duty ratio as a second weight;
the fourth processing unit 24 is connected to the first processing unit 21 and the third processing unit 23, and is configured to calculate a second average value of each first aggregate transaction amount in the first data set and a third average value of each first aggregate transaction amount in the second data set, and perform weighted summation according to the second average value and the first weight, and the third average value and the second weight, to be used as a first aggregate transaction amount of a preset period corresponding to the transaction date, and add the first aggregate transaction amount to the training set.
Specifically, in this embodiment, except for the additional drainage activities, the transaction amounts in the same period of transaction generally have similar trends. And calculating a first average value of first total transaction amounts corresponding to the same transaction date in the same period, and calculating a first duty ratio when the first total transaction amount corresponding to the transaction date is larger than the first average value and a second duty ratio not larger than the first average value, wherein when the first duty ratio is larger than the second duty ratio, the first total transaction amount indicating the transaction date in the same period is higher, and at the moment, the first weight is selected to be larger than the second weight, and the first total transaction amount corresponding to each transaction date corresponding to the first duty ratio is matched with the larger first weight, so that the first total transaction amount finally acquired as training set data is closer to real data. When the first duty cycle is not greater than the second duty cycle, and so on, no further description is provided herein.
In a preferred embodiment of the present invention, the sum of the first weight and the second weight is 1.
In a preferred embodiment of the present invention, the first transaction processing module 2 comprises:
a fifth processing unit 25, configured to calculate a fourth average value of all-day transaction amounts corresponding to the same transaction date in the same period, and add all-day transaction amounts greater than the fourth average value to a third data set, and add all-day transaction amounts not greater than the fourth average value to a fourth data set;
a sixth processing unit 26 connected to the fifth processing unit 25, for calculating a third duty ratio of the number of all-day transaction amounts in the third data set to the number of all-day transaction amounts, and a fourth duty ratio of the number of all-day transaction amounts in the fourth data set to the number of all-day transaction amounts, respectively;
a seventh processing unit 27, connected to the sixth processing unit 26, configured to save a correspondence between a preconfigured duty cycle and a corresponding weight, and match the corresponding weight according to the third duty cycle as a third weight, and match the corresponding weight according to the fourth duty cycle as a fourth weight;
the eighth processing unit 28 is connected to the fifth processing unit 25 and the seventh processing unit 27, and is configured to calculate a fifth average value of all-day transaction amounts in the third data set and a sixth average value of all-day transaction amounts in the fourth data set, and perform weighted summation according to the fifth average value and the third weight, and the sixth average value and the fourth weight, and add the weighted summation to the training set as all-day transaction amounts corresponding to the transaction dates.
Specifically, in this embodiment, except for the additional drainage activities, the transaction amounts in the same period of transaction generally have similar trends. And calculating a fourth average value of all-day transaction amounts corresponding to the same transaction date in the same period, and calculating a third duty ratio when the corresponding all-day transaction amount in the transaction date is larger than the fourth average value and a fourth duty ratio not larger than the fourth average value, wherein when the third duty ratio is larger than the fourth duty ratio, the all-day transaction amount of the transaction date in the same period is higher, and at the moment, the third weight is selected to be larger than the fourth weight, and the larger third weight is matched for the all-day transaction amount corresponding to each transaction date corresponding to the third duty ratio, so that the finally acquired all-day transaction amount serving as training set data is closer to real data. When the third duty cycle is not greater than the fourth duty cycle, and so on, no further description is provided herein.
In a preferred embodiment of the present invention, the sum of the third weight and the fourth weight is 1.
In the preferred embodiment of the present invention, the system further comprises a second transaction processing module 5, connected to the transaction data acquisition module 1, for processing the historical transaction amount data to obtain a first trend curve of the historical transaction amount changing with the transaction period under the transaction date;
the transaction amount prediction module 4 includes:
a similarity calculating unit 41, configured to process a second trend curve of the historical transaction amount of the preset period before the predicted date according to the transaction period, and calculate the total similarity between the second trend curve and the contemporaneous first trend curve;
and a prediction unit 42 connected to the similarity calculation unit 41, for obtaining a service transaction amount prediction result of the predicted date according to the total daily transaction amount of the transaction date corresponding to the total similarity and the predicted total daily transaction amount.
In a preferred embodiment of the present invention, the prediction unit 42 includes:
a storage subunit 421, configured to store at least one pre-configured similarity interval, where each similarity interval corresponds to a set of weight values, and the weight values include at least a fifth weight and a sixth weight;
the calculating subunit 422 is connected with the storage subunit 421, and is configured to obtain a corresponding weight value according to the similarity matching, and perform weighted summation calculation according to the total daily transaction amount and the fifth weight, and the predicted total daily transaction amount and the sixth weight to obtain a service transaction amount prediction result of the predicted date.
In a preferred embodiment of the present invention, the preset time period includes five working days before the predicted date;
the similarity calculation unit 41 includes:
a curve generating subunit 411, configured to generate a second trend curve of the historical transaction amount of each transaction date in the five working days according to the transaction period;
a first calculating subunit 412, a connection curve generating subunit 411, configured to calculate sub-similarities between the second trend curve and the contemporaneous first trend curve, and arrange the sub-similarities in order from big to small to form a similarity queue;
the second calculating subunit 413 is connected to the first calculating subunit 412, and is configured to store five pre-configured weight queues formed by arranging the weight queues in order from large to small, and perform weighted summation on the corresponding sorting positions according to the similarity queues and the weight queues to obtain the total similarity.
In a preferred embodiment of the present invention, the sum of the weights in the weight queue is 1.
In the preferred embodiment of the present invention, the system further comprises a business anomaly detection module 6, which is respectively connected with the transaction data acquisition module 1 and the first transaction processing module 2, and the business anomaly detection module 6 comprises,
an initializing unit 61 that initializes historical business transaction amount data;
the training unit 62 is connected with the initializing unit 61 and learns the statistical rule of the historical business transaction amount data to obtain a rule set with an upper limit index and a lower limit index;
an anomaly detection unit 63, connected to the training unit 62, for determining whether the historical service data amount is within the range of the online index and the offline index of the rule set; and
if so, the historical service transaction line data is normal, otherwise, the historical service transaction line data is abnormal;
the first transaction processing module 2 processes the normal historical business transaction amount data to obtain a training set.
Specifically, in this embodiment, the system needs to perform abnormality detection on the transaction amount, mainly by using a service abnormality detection module to solve the technical problem of abnormality detection on the transaction amount, where the service abnormality detection module may call historical service transaction amount data (hereinafter referred to as transaction data) of a period of time from a core system or a service system, and the service abnormality detection module may apply the obtained mathematical rule to the current transaction data through learning the transaction data of a period of time in the past, so as to determine whether the current transaction data is within a reasonable range, and finally store normal transaction data as historical service transaction amount data of normal transaction, so as to learn a subsequent mathematical rule, thereby playing a role in updating in practice.
In this embodiment, the service anomaly detection module includes an initialization unit, where the initialization unit is configured to initialize 'T0 service normal transaction day data', where the "T0 service normal transaction day data" may be regarded as previous "transaction data", but the "T0 service normal transaction day data" generally appears in a form of a table, and is mainly configured to express a transaction date, a transaction period, and a correspondence between the transaction date, the transaction period, and a historical service transaction amount. The data objects in the table are: t0 business daily time-period transaction amount; the data content is: transaction date, transaction period (0-23), historical transaction amount (corresponding to the date and time), wherein, for example, 0 time may represent 0:00:00-0:59:59, and the following is not exemplified in a one-to-one manner.
The training unit in the business anomaly detection system is connected with the initialization unit and is mainly used for rule training, namely a reference standard set (which can be simply called a rule set) used for detecting business anomalies, wherein the rule training is to learn corresponding statistical rules from transaction data with a selected length and calculate the rule set. This step need not be performed in real time, and can be performed in the early morning 0:00-1:00 is calculated here, but not limited to this time, which is taken as an example.
In a preferred embodiment of the present invention, the training unit includes:
a reading unit for reading the historical business transaction amount data;
the proportion index unit is connected with the reading unit and calculates the proportion index of the historical business transaction amount data of each transaction period according to the date;
the calculating unit is respectively connected with the proportion index unit and the reading unit, calculates the average value and the standard deviation of the proportion index of each transaction period, and calculates the average value and the standard deviation of the historical business transaction amount data of each transaction period;
and the rule set unit is used for obtaining a rule set with an upper limit index and a lower limit index according to the average value and the standard deviation of the proportion index and the average value and the standard deviation of the historical business transaction amount data.
In this embodiment, the step of training the rule of the training unit may include: step 1.1: the data of 5 days closest to the current date is read from the 'T0 business normal transaction date data', and in this embodiment, the data of all or part of other days may be read by taking 5 days as an example. The whole step 1.1 is completed by a reading unit in the training unit, then step 1.2 is executed, step 1.2 is executed by a proportion index unit in the training unit, and step 1.2 mainly comprises calculating a proportion index P of each period according to a trade day (namely the working day), wherein a calculation formula of the proportion index is as follows:
wherein T represents transaction amount, d represents date code, and r represents time period code.
In a preferred embodiment of the present invention, the computing unit includes:
a proportion index calculation unit calculating the average value and standard deviation of the proportion index of each trade period;
and the transaction amount data calculation unit is used for calculating the average value and standard deviation of the historical business transaction amount data of each transaction period.
Step 1.3 and step 1.4 are calculating the mean value and standard deviation of the proportion index of each trade period, and calculating the mean value and standard deviation of the historical business trade amount data of each trade period, which can be completed by a calculating unit, step 1.3, the proportion index calculating unit calculates the proportion index mean PrBAR and standard deviation PrSD of each period based on the proportion index of each trade period in the working day, and the calculation formula is as follows:
step 1.4: the transaction amount data calculation unit calculates the formula of the transaction amount mean value TrBAR and the standard deviation TrSD of each transaction period based on the transaction amount data of each transaction period in the working day as follows:
step 1.5, a rule set unit in the training unit calculates rule sets of each period by using the index data (the average value and standard deviation according to the proportionality index and the average value and standard deviation of the historical business transaction amount data) obtained in the steps 1.1-1.4, wherein the index rule sets have an upper limit U and a lower limit L, and the calculation formula of the rule sets is as follows:
the foregoing description is only illustrative of the preferred embodiments of the present invention and is not to be construed as limiting the scope of the invention, and it will be appreciated by those skilled in the art that equivalent substitutions and obvious variations may be made using the description and drawings, and are intended to be included within the scope of the present invention.
Claims (9)
1. A transaction amount prediction system based on a transaction amount prediction model, comprising:
the transaction data acquisition module is used for acquiring historical service transaction amount data, wherein the historical service transaction amount data comprises a transaction date, a transaction period and historical service transaction amounts corresponding to the transaction date and the transaction period;
the first transaction processing module is connected with the transaction data acquisition module and is used for processing the historical business transaction amount data to obtain a first total transaction amount and an all-day transaction amount which correspond to the transaction date and are in a preset period as a training set;
the prediction model training module is connected with the first transaction processing module and is used for training according to the training set to obtain a transaction amount prediction model taking the transaction date and the first total transaction amount as input and the total transaction amount as output;
the transaction amount prediction module is respectively connected with the transaction data acquisition module and the prediction model training module and is used for inputting a predicted date and a corresponding total transaction amount on the same day into the transaction amount prediction model to obtain a predicted total-day transaction amount, and processing the historical transaction amount data in a preset period before the predicted date and the predicted total-day transaction amount to obtain a service transaction amount prediction result of the predicted date;
the system also comprises a business abnormality detection module which is respectively connected with the transaction data acquisition module and the first transaction processing module, wherein the business abnormality detection module comprises,
an initializing unit for initializing the historical business transaction amount data;
the training unit is connected with the initializing unit and learns the statistical rule of the historical business transaction amount data to obtain a rule set with an upper limit index and a lower limit index;
the anomaly detection unit is connected with the training unit and used for judging whether the historical service data amount is in the range of the upper limit index and the lower limit index of the rule set; and
if so, the historical service transaction amount data is normal, otherwise, the historical service transaction amount data is abnormal;
the first transaction processing module processes the normal historical business transaction limit data to obtain the training set;
the training unit includes:
a reading unit for reading the historical business transaction amount data;
the proportion index unit is connected with the reading unit and calculates the proportion index of the historical business transaction amount data of each transaction period according to the date;
the calculating unit is respectively connected with the proportion index unit and the reading unit, calculates the average value and the standard deviation of the proportion index of each transaction period, and calculates the average value and the standard deviation of the historical business transaction amount data of each transaction period;
the rule set unit is used for obtaining a rule set with an upper limit index and a lower limit index according to the average value and the standard deviation of the proportion index and the average value and the standard deviation of the historical service transaction amount data;
the calculation formula of the proportion index is as follows:
wherein, the liquid crystal display device comprises a liquid crystal display device,
t is used to represent the transaction amount;
d is used to represent a date code;
r is used to represent the period code.
2. The transaction amount prediction system based on a transaction amount prediction model of claim 1, wherein the first transaction processing module comprises:
a first processing unit, configured to calculate a first average value of the first aggregate transaction amounts corresponding to the same transaction date in the same period, and add the first aggregate transaction amount greater than the first average value to a first data set, and add the first aggregate transaction amount not greater than the first average value to a second data set;
a second processing unit, connected to the first processing unit, for calculating a first duty ratio of the number of the first total amount of transactions in the first data set to the number of all the first total amount of transactions, and a second duty ratio of the number of the first total amount of transactions in the second data set to the number of all the first total amount of transactions, respectively;
the third processing unit is connected with the second processing unit and is used for storing the corresponding relation between the pre-configured duty ratio and the corresponding weight, matching the corresponding weight according to the first duty ratio to serve as a first weight, and matching the corresponding weight according to the second duty ratio to serve as a second weight;
and the fourth processing unit is respectively connected with the first processing unit and the third processing unit and is used for calculating a second average value of each first total transaction amount in the first data set and a third average value of each first total transaction amount in the second data set, and adding the third average value and the second weight into the training set as the first total transaction amount of a preset period corresponding to the transaction date according to the second average value and the first weight.
3. The transaction amount prediction system based on a transaction amount prediction model of claim 2, wherein a sum of the first weight and the second weight is 1.
4. The transaction amount prediction system based on a transaction amount prediction model of claim 1, wherein the first transaction processing module comprises:
a fifth processing unit, configured to calculate a fourth average value of the all-day transaction amounts corresponding to the same transaction date in the same period, and add the all-day transaction amounts greater than the fourth average value to a third data set, and add the all-day transaction amounts not greater than the fourth average value to a fourth data set;
a sixth processing unit, connected to the fifth processing unit, configured to respectively calculate a third duty ratio of the number of all-day transaction amounts in the third data set to the number of all-day transaction amounts, and a fourth duty ratio of the number of all-day transaction amounts in the fourth data set to the number of all-day transaction amounts;
a seventh processing unit, connected to the sixth processing unit, configured to save a correspondence between a preconfigured duty ratio and a corresponding weight, and match the corresponding weight according to the third duty ratio as a third weight, and match the corresponding weight according to the fourth duty ratio as a fourth weight;
and an eighth processing unit, connected to the fifth processing unit and the seventh processing unit, respectively, configured to calculate a fifth average value of each of the all-day transaction amounts in the third data set and a sixth average value of each of the all-day transaction amounts in the fourth data set, and perform weighted summation according to the fifth average value and the third weight, and the sixth average value and the fourth weight, to be used as the all-day transaction amount corresponding to the transaction date to be added into the training set.
5. The transaction amount prediction system based on the transaction amount prediction model of claim 4, wherein,
the sum of the third weight and the fourth weight is 1.
6. The transaction amount prediction system based on the transaction amount prediction model according to claim 1, wherein,
the second transaction processing module is connected with the transaction data acquisition module and is used for processing the historical service transaction amount data to obtain a first trend curve of the historical service transaction amount changing along with the transaction period under the transaction date;
the transaction amount prediction module includes:
a similarity calculating unit, configured to process a second trend curve of the historical transaction amount of the preset period before the predicted date according to the transaction period, and calculate a total similarity between the second trend curve and the first trend curve in the same period;
and the prediction unit is connected with the similarity calculation unit and is used for processing the total daily transaction amount of the transaction date corresponding to the total similarity and the predicted total daily transaction amount to obtain a service transaction amount prediction result of the predicted date.
7. The transaction amount prediction system based on a transaction amount prediction model according to claim 6, wherein the prediction unit includes:
the storage subunit is used for storing at least one preset similarity interval, each similarity interval corresponds to a group of weight values, and the weight values comprise at least one fifth weight and one sixth weight;
and the calculation subunit is connected with the storage subunit and is used for obtaining the corresponding weight value according to the similarity matching, and carrying out weighted summation calculation according to the all-day transaction amount and the fifth weight and the predicted all-day transaction amount and the sixth weight to obtain a service transaction amount prediction result of the predicted date.
8. The transaction amount prediction system based on a transaction amount prediction model of claim 6, wherein the preset time period includes five workdays prior to the prediction date;
the similarity calculation unit includes:
a curve generation subunit, configured to generate the second trend curve of the historical transaction amount according to the transaction period for each of the five working days;
the first calculation subunit is connected with the curve generation subunit and is used for respectively calculating the sub-similarity of the second trend curve and the first trend curve in the same period, and arranging the sub-similarity according to the sequence from big to small to form a similarity queue;
the second calculation subunit is connected with the first calculation subunit and is used for storing five pre-configured weight queues which are formed by arranging the weight queues in a sequence from big to small, and carrying out corresponding sequencing position weighted summation according to the similarity queues and the weight queues to obtain the total similarity.
9. The transaction amount prediction system based on a transaction amount prediction model of claim 8, wherein the sum of the weights in the weight queue is 1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110105443.7A CN112785063B (en) | 2021-01-26 | 2021-01-26 | Transaction amount prediction system based on transaction amount prediction model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110105443.7A CN112785063B (en) | 2021-01-26 | 2021-01-26 | Transaction amount prediction system based on transaction amount prediction model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112785063A CN112785063A (en) | 2021-05-11 |
| CN112785063B true CN112785063B (en) | 2023-07-04 |
Family
ID=75757908
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110105443.7A Active CN112785063B (en) | 2021-01-26 | 2021-01-26 | Transaction amount prediction system based on transaction amount prediction model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112785063B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105654381A (en) * | 2015-12-28 | 2016-06-08 | 上海瀚银信息技术有限公司 | Predicting system for business transaction volume |
| CN108288161A (en) * | 2017-01-10 | 2018-07-17 | 第四范式（北京）技术有限公司 | The method and system of prediction result are provided based on machine learning |
| CN109583729A (en) * | 2018-11-19 | 2019-04-05 | 阿里巴巴集团控股有限公司 | Data processing method and device for platform on-time model |
| CN110322085A (en) * | 2018-03-29 | 2019-10-11 | 北京九章云极科技有限公司 | A kind of customer churn prediction method and apparatus |
-
2021
- 2021-01-26 CN CN202110105443.7A patent/CN112785063B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105654381A (en) * | 2015-12-28 | 2016-06-08 | 上海瀚银信息技术有限公司 | Predicting system for business transaction volume |
| CN108288161A (en) * | 2017-01-10 | 2018-07-17 | 第四范式（北京）技术有限公司 | The method and system of prediction result are provided based on machine learning |
| CN110322085A (en) * | 2018-03-29 | 2019-10-11 | 北京九章云极科技有限公司 | A kind of customer churn prediction method and apparatus |
| CN109583729A (en) * | 2018-11-19 | 2019-04-05 | 阿里巴巴集团控股有限公司 | Data processing method and device for platform on-time model |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112785063A (en) | 2021-05-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Whittington | Some basic properties of accounting ratios | |
| CN110135702A (en) | Appraisal procedure, device, system and recording medium are actively spent in a kind of refund of real-time update | |
| CN112785063B (en) | Transaction amount prediction system based on transaction amount prediction model | |
| Sarkar et al. | Testing predictability and nonlinear dependence in the Indian stock market | |
| CN115759751A (en) | Enterprise risk prediction method and device, storage medium, electronic equipment and product | |
| CN115034812A (en) | Steel industry sales prediction method and device based on big data | |
| CN114298825A (en) | Method and device for extremely evaluating repayment volume | |
| Kjellson | Forecasting expected shortfall: an extreme value approach | |
| CN112346995A (en) | Construction method and device of test risk estimation model based on banking industry | |
| Fioramanti et al. | The Econometrics of the EU Fiscal Governance: is the European Commission methodology still adequate? | |
| Lei et al. | Analysis of gold and bitcoin price prediction based on LSTM model | |
| CN111640026A (en) | Financial multi-factor prediction method based on sparse non-negative linear regression | |
| Smith et al. | A method of parameterising a feed forward multi-layered perceptron artificial neural network, with reference to South African financial markets | |
| CN115994679B (en) | Regional power grid active planning method and system based on load prediction correction | |
| CN116029818A (en) | Method, device, equipment and storage medium for determining risk value | |
| CN117236718A (en) | Power grid planning determination method and device, electronic equipment and medium | |
| CN114897381A (en) | Accounting evaluation method, device, equipment, medium and product | |
| Liu et al. | LSTM-based cross-prediction price model for gold and bitcoin | |
| O’Grady | Estimating the Output, Inflation and Unemployment Gaps in Ireland using Bayesian Model Averaging | |
| CN115170329A (en) | Investment benefit evaluation method for science and technology project | |
| CN114638687A (en) | Interest rate prediction method, device, electronic equipment and storage medium | |
| Elsaied | Developing and Forecasting the Egyptian Construction Cost Index | |
| CN117495554A (en) | Option buying and selling time determining system and method based on time convolution network | |
| CN115034909A (en) | Product configuration method, device, equipment and storage medium based on LSTM model | |
| CN115860562A (en) | Software workload rationality evaluation method, device and equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |