TW201939315A - Service occurrence amount prediction method, apparatus and device - Google Patents

Abstract

A service occurrence amount prediction method, apparatus and device. The method comprises: discretizing historical service data prior to a predetermined time period to obtain a service occurrence amount vector of time granularity, and generating a service occurrence amount distribution feature vector according to the service occurrence amount in continuous time in the historical service data; and finally, determining, according to the service occurrence amount vector of time granularity and the service occurrence amount distribution feature vector, a service occurrence amount within the determined time period.

Description

Method, device and equipment for predicting business volume

This specification relates to the field of computer technology, and in particular, to a method, device, and equipment for predicting the amount of business occurrence.

With the continuous development of network technology and terminal technology, e-commerce is becoming more and more important in people's daily lives. For example, people can buy various goods in shopping websites through online payment. In addition, businesses such as online overseas purchases and offline face-to-face payments have also developed rapidly. In this way, payment applications (such as Alipay) need to support merchants and buyers to make payments and receive payments in different currencies. In this way, the payment application needs to settle the currency of the corresponding country to the corresponding merchant. Therefore, the payment application has a large amount of exchange requirements.
Generally, payment applications need to purchase a certain amount of foreign exchange every working day to meet business needs. In order to minimize the impact of exchange rate fluctuations on payment applications, the payment application locks the amount of foreign exchange purchases on the day with the counterparty on the working day. In practical applications, the development trend in a predetermined time period can be given through time series algorithms, such as moving average method, moving average, ARIMA (Autoregressive Integrated Moving Average Model) or Holt-Winters. However, the above-mentioned time series algorithm requires higher consistency of time series trends. If the recent business development trend is abnormal, the prediction results obtained based on the above algorithm are likely to be abnormal, resulting in a deviation in the predicted value. In this way, it is necessary to provide a solution that can accurately and instantly predict the amount of business occurrence, and can reduce business risks and improve capital utilization efficiency.

The purpose of the embodiments of the present specification is to provide a method, an apparatus, and a device for predicting the amount of business occurrences, so as to provide a solution that can accurately and immediately predict the amount of business occurrences, and at the same time, can reduce business risks and improve capital utilization efficiency.
In order to implement the above technical solution, the embodiments of this specification are implemented as follows:
A method for predicting the amount of service occurrence provided in the embodiments of this specification, the method includes:
Discretizing historical business data before a predetermined period of time to obtain a time-granularity business volume vector, and generating a business volume distribution feature vector according to continuous-time business volume in the historical business data;
Determining the amount of business occurrences within the predetermined time period according to the time-granularity business occurrence vector and the business occurrence distribution feature vector.
Optionally, the historical service data includes a first historical service data in a first time period closest to the predetermined time period and a second historical service data other than the first historical service data,
The generating a feature vector of a service occurrence distribution according to the service occurrence vector of the time granularity and the continuous occurrence of the service occurrence in the historical service data includes:
Generating a first feature vector of the distribution of service occurrences according to the continuous service occurrences in the first historical service data;
And generating a second feature vector of the distribution of business occurrences according to the continuous business occurrences in the second historical business data.
Optionally, before determining the service occurrence volume within the predetermined time period according to the service occurrence vector and the service occurrence distribution feature vector of the time granularity, the method further includes:
Determining a similarity between the first feature vector and the second feature vector;
Determine the weight of the second feature vector according to the similarity between the first feature vector and the second feature vector.
Optionally, determining the similarity between the first feature vector and the second feature vector includes:
The similarity between the first feature vector and the second feature vector is determined by any of the following methods: Euclidean distance, the cosine of the angle of the vector, and the absolute value of the difference of the vector.
Optionally, the determining the service occurrence amount within the predetermined time period according to the service occurrence vector and the service occurrence distribution feature vector of the time granularity includes:
Merging the time-granularity business volume vector with the first feature vector and the second feature vector to obtain a combined first feature vector and a second feature vector;
Based on the second feature vector and the weight of the second feature vector, optimizing initial parameters through a loss function and a predetermined parameter optimization algorithm to obtain optimized initial parameters;
Determine the amount of traffic generated in the predetermined time period according to the optimized initial parameters and the first feature vector.
Optionally, the predetermined parameter optimization algorithm includes a gradient descent algorithm, a Newton method, a quasi-Newton method, a conjugate gradient method, and a heuristic optimization algorithm.
An apparatus for predicting the amount of traffic generated in the embodiments of the present specification, the apparatus includes:
A processing module, configured to discretize historical business data before a predetermined period of time, to obtain a time granularity business volume vector, and to generate a business volume distribution characteristic based on the continuous business volume in the historical business data vector;
The traffic occurrence forecasting module is configured to determine a traffic occurrence in the predetermined time period based on the time granularity traffic occurrence vector and the traffic occurrence distribution feature vector.
Optionally, the historical service data includes a first historical service data in a first time period closest to the predetermined time period and a second historical service data other than the first historical service data,
The processing module includes:
A first feature vector generating unit, configured to generate a first feature vector of a service volume distribution according to a service volume of continuous time in the first historical service data;
A second feature vector generating unit is configured to generate a second feature vector of a service volume distribution according to the service volume of continuous time in the second historical service data.
Optionally, the apparatus further includes:
A similarity determination module, configured to determine a similarity between the first feature vector and the second feature vector;
A weight determining module is configured to determine a weight of the second feature vector according to a similarity between the first feature vector and the second feature vector.
Optionally, the similarity determination module is configured to determine the similarity between the first feature vector and the second feature vector through any of the following devices: Euclidean distance, angle cosine of the vector, and vector The absolute value of the difference.
Optionally, the traffic occurrence prediction module includes:
A merging unit for merging the time-granularity business volume vector with the first feature vector and the second feature vector to obtain a merged first feature vector and a second feature vector;
An initial parameter optimization unit, configured to optimize the initial parameters based on the second feature vector through a loss function and a predetermined parameter optimization algorithm to obtain optimized initial parameters;
A traffic occurrence prediction unit is configured to determine a traffic occurrence in the predetermined time period according to the optimized initial parameters and the first feature vector.
Optionally, the predetermined parameter optimization algorithm includes a gradient descent algorithm, a Newton method, a quasi-Newton method, a conjugate gradient method, and a heuristic optimization algorithm.
An embodiment of this specification provides a device for predicting the amount of traffic generated. The device for predicting the amount of traffic generated includes:
A processor; and a memory arranged to store computer-executable instructions that, when executed, cause the processor to:
Discretizing historical business data before a predetermined period of time to obtain a time-granularity business volume vector, and generating a business volume distribution feature vector according to continuous-time business volume in the historical business data;
Determining the amount of business occurrences within the predetermined time period according to the time-granularity business occurrence vector and the business occurrence distribution feature vector.
As can be seen from the technical solutions provided by the embodiments of the present specification, the embodiments of the present specification obtain a time-granularity business volume vector by discretizing historical business data before a predetermined period of time. The amount of business occurrences over time generates a feature vector of business occurrences distribution. Finally, the amount of business occurrences in a predetermined period of time can be determined based on the time-granularity business occurrence vector and business occurrence distribution feature vector. Vector (which can include multi-dimensional feature vectors) to predict the business volume in a predetermined period of time, and the final forecast result can be adjusted accordingly when the short-term business volume fluctuates, effectively avoiding large business volume fluctuations in the short-term The situation that affects the final prediction result occurs. In addition, by using the feature vector of the business volume distribution, it is also possible to better capture the change trend of the business volume in a specified time period (such as the business volume change trend in historical business data). Etc.) to improve business The accuracy of the forecast of the occurrence amount reduces business risks and improves the efficiency of capital utilization.

The embodiments of the present specification provide a method, a device, and a device for predicting the amount of traffic generated.
In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described in combination with the drawings in the embodiments of this specification. Obviously, the described The examples are only a part of examples of this specification, but not all examples. Based on the embodiments in this specification, all other embodiments obtained by a person of ordinary skill in the art without creative efforts should fall within the protection scope of this specification.

Example one
As shown in FIG. 1, an embodiment of the present specification provides a method for predicting the amount of traffic generated. The method may be executed by a terminal device or a server. The terminal device may be a personal computer or a mobile phone or a tablet. A mobile terminal device such as a computer. The terminal device may be a terminal device used by a user. The server may be an independent server or a server cluster composed of multiple servers. This method can be used for accurate real-time prediction of the occurrence of traffic and the like. In this embodiment, a server is taken as an example for description. For the situation of a terminal device, it can be processed according to the following related content, which will not be repeated here. The method may specifically include the following steps:
In step S102, the historical service data before a predetermined time period is discretized to obtain a time-granularity service volume vector.
The predetermined time period can be set according to the actual situation. For example, the predetermined time period can be a certain time in the future, such as from the current time point to a time point after an interval of 1 hour, etc., or it can be a time before the current time point. Paragraph etc. The discretization process can be a method of mapping the limited individuals in infinite space to a limited space to improve the space-time efficiency of the algorithm. That is, the discretization process is performed without changing the relative size of the data. The corresponding reduction method of the data. The discretization process is large in the data itself, and it cannot save the corresponding attributes as the index of the array. If only the relative attributes of the data are needed at this time, the data can be processed. Discretization processing, that is, when the data is only related to the relative size between them, and has nothing to do with the specific content of the data, the data can be discretized. The traffic volume vector may be a vector of related data generated by a user when completing one or more services. The business volume can be the number of users who have performed or completed one or more services, or the number of services that users have performed or completed, for example, as shown in Figure 2, which can include multiple servers, and different servers can To provide services for different businesses, people can complete different businesses through different business servers. The amount of business occurrences can include multiple types, such as transaction volume and conversion volume. Time granularity can be the basic time unit for business occurrence or business volume statistics, such as 10 minutes or 15 minutes.
In the implementation, due to the rapid development of online overseas purchases and offline face-to-face payment services, payment applications (such as Alipay, etc.) support merchants and buyers to make payments and receive payments in different currencies. If the merchant and the buyer (that is, the user) belong to different countries or regions, then whether online or offline, the user only needs to use the local currency to pay. In this way, the payment application needs to settle the corresponding foreign currency to different merchants. Therefore, there is a large amount of exchange requirements for payment applications. In this way, payment applications need to purchase a certain amount of foreign exchange every working day to meet business needs. In order to minimize the impact of exchange rate fluctuations on the payment application, the payment application locks the amount of foreign exchange purchases on the day with the counterparty on the working day. In this way, the payment application needs to purchase a certain amount of foreign exchange on each working day to meet business needs. An accurate and real-time method can predict business volume, reduce business risks, and improve capital utilization efficiency. In practical applications, time series algorithms can be used, such as moving average, moving average, ARIMA, or Holt-Winters. The input parameters are time series values, and the development of the predetermined time period is given according to different algorithms. However, the above-mentioned time series algorithm requires higher consistency of time series trends, that is, if the recent business development trend is abnormal, the prediction results obtained based on the above algorithm are also likely to be abnormal, and the above time The sequence algorithm has a low utilization rate of real-time data. In practice, it cannot accurately capture the change trend of the latest business volume on the day, which causes a large deviation in the predicted value.
In addition, you can also use the ratio method to predict the business generation in real time. This method calculates the total business generation for the day by using the ratio of the business generation per hour and the real-time business generation in the hour. The method requires high consistency in the proportion of the hourly service generation. If the hourly service generation changes suddenly, the prediction result obtained by the method is likely to be abnormal.
It can be seen that, in order to accurately capture the latest trend of business occurrences on the day and reduce prediction errors, it can be handled in the following ways, which can specifically include the following:
You can first determine the time period that needs to be predicted. For example, from the current time point to the time point after 2 hours interval. For example, if the current time point is 10 o'clock and the time point after 2 hours is 12 o'clock, you need to predict the time period. (That is, the predetermined time period) can be 10 to 12 o'clock; for another example, the current time point is 10 o'clock, and the time period to be predicted (that is, the predetermined time period) can be 10 o'clock to 24 o'clock on the day, etc. . The server may store related data of one or more services, and the data may be related data generated after the user requests a related service and completes the service. After the server determines the predetermined time period, it can extract a certain length of time (such as before 10 o'clock in the current time) from the corresponding business data stored before the predetermined time period (such as 10 o'clock in the current time to 24 o'clock in the current day). One month or two months, etc.). Generally, the historical business data has a large amount of data. In order to reduce the processing pressure of the server, the obtained historical business data may be discretized. Through the discretization of historical business data, historical business data can be discretized into time-granularity business volume vectors according to predetermined rules. Among them, historical business data can be divided into two parts, and one part can be from the current day. The historical business data from the latest business occurrence time (that is, the current time point) in the real-time business data from zero to the current business data. This part of the historical business data can be discretized into the business volume vector according to the above rules. The other part can be the historical business before zero Data, this part of the data can also be divided into two parts, one of which can be the historical business data from the zero point of each day to the time corresponding to the current time point in the day, for example, the current time point is 10 o'clock on the day Date is February 28th, the historical business data of this part can include historical business data from February 27th to zero o'clock, February 10th, and historical business data from February 26th to 10 o'clock, February 25th. Historical business data from 0 to 10 o'clock, and so on; the other part can be the time corresponding to the current time point in each day Moments ~ Historical business data at 24 o'clock on that day. For example, based on the above example, historical business data in this part may include historical business data from 10 o'clock to 24 o'clock on February 27, and 10 o'clock on February 26 ~ Historical business data at 24 o'clock, historical business data from 10 o'clock to 24 o'clock on February 25, and so on. The historical business data of each of the above parts can be discretized into a business occurrence vector according to the above rules.
It should be noted that the total amount of business occurrences that can be summed up at the remaining time of the day (that is, 10 o'clock in the current time to 24 o'clock in the current day), that is, the amount of business that needs to be predicted.
In step S104, a service occurrence quantity distribution feature vector is generated according to the continuous occurrence service amount in the historical service data.
In implementation, a rule or algorithm for generating a feature vector may be set in the server in advance, and the corresponding rule or algorithm may be determined according to actual conditions, which is not limited in the embodiments of the present specification. The server can first extract continuous business volume from the historical business data obtained, and can use the rules or algorithms set above to generate corresponding characteristics of continuous business volume in historical business data to characterize the distribution of business volume. The generated feature may be a distribution feature. Then, the server may combine the obtained distribution feature with the time-granularity business volume vector obtained in step S102 to obtain a business volume distribution feature vector. Among them, the characteristic vector of the distribution of business occurrences can have various expressions, such as the average business occurrence per hour, the increase of the business occurrence per hour, and the growth rate of the business occurrence per hour.
In actual applications, in addition to the above-mentioned commonly used business volume distribution feature vectors, other feature vectors may be included, such as the periodic business volume distribution feature vector (specifically, such as the business volume distribution feature vector per week, the quarterly business volume Distribution feature vector, distribution feature vector of business volume for each month), distribution feature vector of activity business distribution (specifically, the distribution pattern vector of business volume on the promotion day), moving characteristic distribution feature vector of moving average (specificity, such as calculation Distribution of the average business volume within a certain period of the window).
In step S104, the service occurrence amount within a predetermined time period is determined according to the service occurrence vector and the service occurrence distribution feature vector of the time granularity.
In implementation, the specific process of determining the business volume in a predetermined time period according to the business granularity vector and the business volume distribution feature vector of the time granularity described above can be implemented in various ways, in order to be well reflected in historical business data. The outbound business volume trend (or the past trend) can be used to fit the distribution of the business volume trend in a certain period of time (that is, a predetermined period of time). A dictionary learning algorithm can be used to implement the The forecast of the business volume of the dictionary, the dictionary learning algorithm can include two phases, namely the dictionary construction phase and the use of a dictionary to represent the business volume phase, each of the above two phases can be achieved through many different algorithms. Dictionary learning is essentially a dimensionality reduction representation of a huge data set. In addition, dictionary learning always tries to learn the most rustic features hidden in the depth of historical business data. Dictionary learning can be a linear representation of the business volume distribution feature vector and initial parameters used to predict the business volume. The initial parameters can be obtained through continuous iterative optimization based on historical business data through loss functions and parameter optimization algorithms. The loss function can be used to measure the degree of fit between the above-mentioned business occurrence distribution feature vector and the linear function of the initial parameters. The loss value and gradient of the predicted business occurrence and historical actual business occurrence can be calculated based on the existing linear function. When the loss function is the smallest, it means that the fitting degree is the best, and the corresponding initial parameters are the optimal parameters. The minimum value of the loss function can be solved by the parameter optimization algorithm. The parameter optimization algorithm may include multiple types. For example, a gradient descent method may be specifically used to obtain an optimal parameter. First, you can determine the gradient of the loss function at the current position (the value of the gradient can be calculated from the loss function). You can use the initial step size multiplied by the gradient of the loss function to get the distance that the current position drops. At this time, you can determine whether For all coefficients, the distance of the gradient descent is less than the set error value. If the distance of the gradient descent is less than the set error value, the parameter optimization algorithm is terminated. At this time, the initial parameter is already the optimal parameter. If it is larger than the set error value, all the initial parameters can be updated and the above iterative calculation is continued until the distance of the gradient descent is less than the set error value.
The server can combine the time-granularity business volume vector and the business volume distribution feature vector obtained above by using a preset processing rule or algorithm to obtain a multi-dimensional vector combination related to the predicted business volume. The feature vector of the business volume distribution, the combined feature vector of the business volume distribution may be a feature vector related to the forecast of the business volume. The server can train the above-mentioned loss function and parameter optimization algorithm respectively based on historical business data, determine the relevant parameters in the loss function and parameter optimization algorithm, and then obtain the trained loss function and parameter optimization algorithm. Then, it can be based on After training, the loss function and parameter optimization algorithm continuously and iteratively optimize the initial parameters to obtain the optimal parameters. Then, a linear regression may be performed using the above-mentioned business occurrence distribution feature vector (the business occurrence distribution feature vector corresponding to the historical business data corresponding to the historical business data of the latest business occurrence time in the real-time business data from the current day) and the optimal parameters obtained above may be used for linear regression , You can get the business volume in a predetermined period of time.
It should be noted that, in actual applications, the processing of predicting the amount of traffic generated within a predetermined period of time can be implemented not only through dictionary learning algorithms, but also through other algorithms, such as related algorithms for sparse models. This embodiment of the present specification does not limit this.
The embodiment of the present specification provides a method for predicting the amount of business occurrence. By discretizing historical business data before a predetermined period of time, a time-granularity business occurrence vector can be obtained. In addition, continuous-time business data in historical business data can also be used. Occurrence, generating the business occurrence distribution feature vector. Finally, the business occurrence amount within a predetermined time period can be determined according to the time granularity business occurrence vector and the business occurrence distribution feature vector. In this way, the business occurrence distribution feature vector (may (Including multi-dimensional feature vectors) to predict the business volume in a predetermined period of time, and the final forecast result can be adjusted accordingly when the short-term business volume fluctuates greatly, effectively preventing the business volume from affecting the final short-term fluctuation. The situation of the prediction result occurs, and by using the feature vector of the business volume distribution, the change trend of the business volume (such as the trend of the business volume in the historical business data) can be better captured. This can improve the forecast of business volume. Measurement accuracy, reducing business risks and improving capital utilization efficiency.

Example two
As shown in FIG. 3, an embodiment of the present specification provides a method for predicting the amount of traffic generated. The method may be executed by a terminal device or a server. The terminal device may be a personal computer or a mobile phone or a tablet. A mobile terminal device such as a computer. The terminal device may be a terminal device used by a user. The server may be an independent server or a server cluster composed of multiple servers. This method can be used for accurate real-time prediction of the occurrence of traffic and the like. In this embodiment, a server is taken as an example for description. For the situation of a terminal device, it can be processed according to the following related content, which will not be repeated here. The method may specifically include the following steps:
In step S302, the historical service data before a predetermined time period is discretized to obtain a time-granularity service volume vector.
In implementation, the server can record the relevant business data generated by various services in real time. The recorded related business data can include two parts, one is the business data generated on the day (that is, from the zero point of the day to the current time), and the other part can be It is historical business data before that day. The server can obtain the business data of the above two parts, and can perform data cleaning on the business data of the above two parts according to certain rules. Through the data cleaning, the above two parts of the business data can be re-examined and verified, so that the duplicate information in the two parts of the business data can be deleted, and errors can be corrected. After the corrected business data is obtained, it can be discretized to obtain a time-granularity business volume vector. The specific process of discretizing historical business data to obtain a time-granularity business volume vector can be found above. The relevant content in step S102 in the first embodiment will not be repeated here.
Based on the above, the server may include two parts of historical business data. Therefore, the historical business data may include the first historical business data for the first time period closest to the predetermined time period and the first historical business data other than the first historical business data. Second historical business data, where the first time period closest to the predetermined time period may refer to a time period starting from zero on the current day to the current time point, the second historical business data may be historical business data before the current day, and the second historical business The data can also include two parts, one of which can be the historical business data of each day from the zero point to the time corresponding to the current time point in that day, and the other part can be the time corresponding to the current time point in each day to 24 hours of the day Bell's historical business profile. For the historical service data of the multiple divided sections, the following steps S304 to S316 may be performed respectively.
In step S304, a first feature vector of the service volume distribution is generated according to the service volume for continuous time in the first historical service data.
In implementation, a rule or algorithm for generating a feature vector may be set in the server in advance, and the corresponding rule or algorithm may be determined according to actual conditions, which is not limited in the embodiments of the present specification. The server may first extract continuous service occurrences from the obtained first historical service data, and may use the rules or algorithms set above to generate continuous service occurrences in the first historical service data to generate the first A feature vector. For example, if the first historical business data is historical business data starting from zero on the current day to the current time point, based on the time-granularity business volume vector of the first historical business data, a first feature vector of 1 × N dimensions can be obtained.
In step S306, a second feature vector of the service volume distribution is generated according to the continuous service volume in the second historical service data.
In the implementation, the server may first extract continuous service occurrences from the second historical service data obtained, and may generate continuous service occurrences from the second historical service data to generate service occurrences through the rules or algorithms set above. The second eigenvector of the quantity distribution. For example, for the historical business data in the second historical business data starting from zero on each day to the time corresponding to the current time point in that day, based on the time granularity of the historical business data, the business volume vector can be obtained 1 × N dimensions The second feature vector of the historical business data from the time corresponding to the current time point in each day in the second historical business data to 24 o'clock on that day, the business volume vector based on the time granularity of the historical business data can be obtained 1 × N-dimensional second feature vector.
In step S308, the time-granularity business volume vector is combined with the first feature vector and the second feature vector, respectively, to obtain a combined first feature vector and a second feature vector.
In implementation, the server may combine the first feature vector and the second feature vector obtained above with the time-granularity business volume vector obtained in step S302 above to obtain a combined business volume distribution feature vector (i.e., (Merged first feature vector and second feature vector). Among them, the first feature vector and the second feature vector may have multiple expressions, for example, an average hourly service occurrence amount, an increase in an hourly service occurrence amount, and an increase rate of an hourly service occurrence amount.
It should be noted that the above-mentioned combined first feature vector and second feature vector may include, for example, the average business volume per hour, the increase in the volume of business per hour, and the growth rate of the volume of business per hour. Including, for example, the distribution feature vector of the periodic business occurrence amount, the distribution feature vector of the active business occurrence amount, the moving feature distribution vector of the moving average business occurrence, and the like.
In order to indicate the important relationship between different feature vectors, corresponding weights may be set for the first feature vector and the second feature vector. For details, refer to the processing in steps S310 and S312 described below.
In step S310, a similarity between the first feature vector and the second feature vector is determined.
In implementation, the weights of the first feature vector and the second feature vector can be used to determine the value of the two weights. To this end, a related algorithm for calculating the similarity can be set. In practical applications, you can use any of the following A method determines the similarity between the first eigenvector and the second eigenvector: the Euclidean distance, the cosine of the angle of the vector, and the absolute value of the difference of the vector. Taking the Euclidean distance as an example, the server can calculate the Euclidean distance between the first eigenvector and the second eigenvector. Through the obtained Euclidean distance data, the similarity between the first eigenvector and the second eigenvector can be determined. That is, when the Euclidean distance between the first feature vector and the second feature vector is smaller, it indicates that the closer the first feature vector and the second feature vector are, the higher the similarity between the two, and the subsequent business occurs When predicting the quantity, the proportion of the second feature vector is greater; when the Euclidean distance between the first feature vector and the second feature vector is larger, it indicates that the difference between the first feature vector and the second feature vector is greater ( Or the greater the difference), the lower the similarity between the two, and the smaller the proportion of the second feature vector when the subsequent business volume forecast is performed. Correspondingly, the specific processing method and judgment method of the similarity calculation algorithm based on the angle cosine of the vector or the absolute value of the difference of the vector can be similar to the above-mentioned Euclidean distance. For details, refer to the process of Euclidean distance. This will not be repeated here.
In step S312, the weight of the second feature vector is determined according to the similarity between the first feature vector and the second feature vector.
In implementation, after the server obtains the similarity between the first feature vector and the second feature vector through the processing in step S310, the weight of the second feature vector can be determined based on the similarity between the two. In practical applications, the weight of the second feature vector can be determined. The calculated Euclidean distance between the first feature vector and the second feature vector is used as the weight of the second feature vector, or the cosine of the angle between the first feature vector and the second feature vector may be used as the first The weight of the two feature vectors, or the absolute value of the difference between the first feature vector and the second feature vector may be used as the weight of the second feature vector.
A dictionary learning algorithm can be used to determine the amount of business in a predetermined period of time. The dictionary learning algorithm can include initial parameters, and the dictionary learning algorithm also involves a loss function and parameter optimization algorithm. The specific processing process can be See the processing of steps S314 and S316 below.
In step S314, based on the weights of the second feature vector and the second feature vector, the initial parameters are optimized through a loss function and a predetermined parameter optimization algorithm to obtain optimized initial parameters.
Among them, the predetermined parameter optimization algorithm may include multiple types, and a variety of optional algorithms are provided below, including gradient descent algorithm, Newton method, quasi-Newton method, conjugate gradient method, and heuristic optimization algorithm. The loss function can refer to a function that maps an event to a real number that expresses the economic or opportunity cost associated with the event.
In implementation, the specific content of the loss function and the parameter optimization algorithm can be set in the server in advance. For example, a gradient descent algorithm or Newton's method can be set as the parameter optimization algorithm, and the specific form of the preset loss function can be written. Into the server. The server can iteratively calculate the initial parameters in the dictionary learning algorithm through the weights of the second feature vector and the second feature vector corresponding to the second historical business data, and the preset loss function and parameter optimization algorithm. After multiple iterations, the optimal initial parameters are obtained, and then the optimized initial parameters are obtained.
In step S316, the amount of traffic generated in a predetermined period of time is determined according to the optimized initial parameters and the first feature vector.
For the specific processing procedure of the above step S316, reference may be made to the related content in step S104 in the foregoing first embodiment, and details are not described herein again.
Through the above-mentioned process of steps S302 to S316, the multi-dimensional feature vector is used to predict the amount of business occurrence. In this process, the consistency of the time trend is not high. Large (or, the rise or fall is obvious), this process can automatically adjust the final prediction result; in addition, this process can better capture the change trend of business volume in a specified time period, and learn the calculation through the dictionary The method can well fit the trend distribution of the real-time data from the changing trend of business volume in historical business data. By using this process, the accuracy of real-time prediction is better than that of time series prediction.
The embodiment of the present specification provides a method for predicting the amount of business occurrence. By discretizing historical business data before a predetermined period of time, a time-granularity business occurrence vector can be obtained. In addition, continuous-time business data based on historical business data Occurrence, generating the business occurrence distribution feature vector. Finally, the business occurrence amount within a predetermined time period can be determined according to the time granularity business occurrence vector and the business occurrence distribution feature vector. In this way, the business occurrence distribution feature vector (may (Including multi-dimensional feature vectors) to predict the business volume in a predetermined period of time, and the final forecast result can be adjusted accordingly when the short-term business volume fluctuates greatly, effectively preventing the business volume from affecting the final short-term fluctuation. The situation of the prediction result occurs, and by using the feature vector of the business volume distribution, the change trend of the business volume (such as the trend of the business volume in the historical business data) can be better captured. This can improve the forecast of business volume. Measurement accuracy, reducing business risks and improving capital utilization efficiency.

Example three
The above is a method for predicting the amount of traffic generated by the embodiment of the present specification. Based on the same idea, the embodiment of the present specification also provides a device for predicting the amount of traffic generated, as shown in FIG. 4.
The device for predicting the occurrence of business includes a processing module 401 and a prediction module for business occurrence 402, of which:
The processing module 401 is configured to discretize historical business data before a predetermined period of time, to obtain a time granularity business volume vector, and to generate a business volume distribution based on the continuous business volume in the historical business data. Feature vector;
The traffic occurrence prediction module 402 is configured to determine a traffic occurrence within the predetermined time period according to the time granularity traffic occurrence vector and the traffic occurrence distribution feature vector.
In the embodiment of the present specification, the historical service data includes a first historical service data in a first time period closest to the predetermined time period and a second historical service data other than the first historical service data.
The processing module 401 includes:
A first feature vector generating unit, configured to generate a first feature vector of a service volume distribution according to a service volume of continuous time in the first historical service data;
A second feature vector generating unit is configured to generate a second feature vector of a service volume distribution according to the service volume of continuous time in the second historical service data.
In the embodiment of the present specification, the device further includes:
A similarity determination module, configured to determine a similarity between the first feature vector and the second feature vector;
A weight determining module is configured to determine a weight of the second feature vector according to a similarity between the first feature vector and the second feature vector.
In the embodiment of the present specification, the similarity determination module is configured to determine the similarity between the first feature vector and the second feature vector through any one of the following devices: Euclidean distance, and the angle cosine of the vector, And the absolute value of the difference of the vectors.
In the embodiment of the present specification, the traffic occurrence prediction module 402 includes:
A merging unit for merging the time-granularity business volume vector with the first feature vector and the second feature vector to obtain a merged first feature vector and a second feature vector;
An initial parameter optimization unit, configured to optimize the initial parameters based on the second feature vector and the weight of the second feature vector through a loss function and a predetermined parameter optimization algorithm to obtain optimized initial parameters;
A traffic occurrence prediction unit is configured to determine a traffic occurrence in the predetermined time period according to the optimized initial parameters and the first feature vector.
In the embodiment of the present specification, the predetermined parameter optimization algorithm includes a gradient descent algorithm, a Newton method, a quasi-Newton method, a conjugate gradient method, and a heuristic optimization algorithm.
The embodiment of the present specification provides a device for predicting the amount of traffic generated by discretizing historical service data before a predetermined period of time to obtain a time-granularity service volume vector. In addition, the service can also be based on continuous-time services in the historical service data. Occurrence, generating the business occurrence distribution feature vector. Finally, the business occurrence amount within a predetermined time period can be determined according to the time granularity business occurrence vector and the business occurrence distribution feature vector. In this way, the business occurrence distribution feature vector (may (Including multi-dimensional feature vectors) to predict the business volume in a predetermined period of time, and the final forecast result can be adjusted accordingly when the short-term business volume fluctuates greatly, effectively preventing the business volume from affecting the final short-term fluctuation. The situation of the prediction result occurs, and by using the feature vector of the business volume distribution, the change trend of the business volume (such as the trend of the business volume in the historical business data) can be better captured. This can improve the forecast of business volume. Measurement accuracy, reducing business risks and improving capital utilization efficiency.

Embodiment 4
The above is a device for predicting the amount of traffic generated by the embodiment of the present specification. Based on the same idea, the embodiment of the present specification also provides a device for predicting the amount of traffic generated, as shown in FIG. 5.
The device for predicting the amount of service occurrence may be a server or a terminal device provided in the foregoing embodiment.
The forecasting device for business volume may vary greatly due to different configurations or performance. It may include one or more processors 501 and memory 502. The memory 502 may store one or more storage applications or data. . The memory 502 may be temporarily stored or persistently stored. The application program stored in the memory 502 may include one or more modules (not shown), and each module may include a series of computer-executable instructions in a device for predicting the amount of traffic generated. Furthermore, the processor 501 may be configured to communicate with the memory 502 and execute a series of computer-executable instructions in the memory 502 on a device for predicting the amount of traffic generated. The traffic generation prediction device may further include one or more power sources 503, one or more wired or wireless network interfaces 504, one or more input / output interfaces 505, and one or more keyboards 506.
Specifically, in this embodiment, the device for predicting the amount of traffic generated includes a memory and one or more programs. One or more programs are stored in the memory, and one or more programs may include one or more programs. Modules, and each module may include a series of computer-executable instructions in a device for predicting business volume, and configured to be executed by one or more processors. The one or more programs include the following computers. Executable instructions:
Discretizing historical business data before a predetermined period of time to obtain a time-granularity business volume vector, and generating a business volume distribution feature vector according to continuous-time business volume in the historical business data;
Determining the amount of business occurrences within the predetermined time period according to the time-granularity business occurrence vector and the business occurrence distribution feature vector.
Optionally, the historical service data includes a first historical service data in a first time period closest to the predetermined time period and a second historical service data other than the first historical service data,
The generating a feature vector of a service occurrence distribution according to the service occurrence vector of the time granularity and the continuous occurrence of the service occurrence in the historical service data includes:
Generating a first feature vector of the distribution of service occurrences according to the continuous service occurrences in the first historical service data;
And generating a second feature vector of the distribution of business occurrences according to the continuous business occurrences in the second historical business data.
Optionally, before determining the service occurrence volume within the predetermined time period according to the service occurrence vector and the service occurrence distribution feature vector of the time granularity, the method further includes:
Determining a similarity between the first feature vector and the second feature vector;
Determine the weight of the second feature vector according to the similarity between the first feature vector and the second feature vector.
Optionally, determining the similarity between the first feature vector and the second feature vector includes:
The similarity between the first feature vector and the second feature vector is determined by any of the following methods: Euclidean distance, the cosine of the angle of the vector, and the absolute value of the difference of the vector.
Optionally, the determining the service occurrence amount within the predetermined time period according to the service occurrence vector and the service occurrence distribution feature vector of the time granularity includes:
Merging the time-granularity business volume vector with the first feature vector and the second feature vector to obtain a combined first feature vector and a second feature vector;
Based on the second feature vector and the weight of the second feature vector, optimizing initial parameters through a loss function and a predetermined parameter optimization algorithm to obtain optimized initial parameters;
Determine the amount of traffic generated in the predetermined time period according to the optimized initial parameters and the first feature vector.
Optionally, the predetermined parameter optimization algorithm includes a gradient descent algorithm, a Newton method, a quasi-Newton method, a conjugate gradient method, and a heuristic optimization algorithm.
The embodiment of the present specification provides a device for predicting the amount of traffic generated by discretizing historical service data before a predetermined period of time to obtain a time-granularity service volume vector. In addition, the service can also be based on continuous-time services in historical service data Occurrence, generating the business occurrence distribution feature vector. Finally, the business occurrence amount within a predetermined time period can be determined according to the time granularity business occurrence vector and the business occurrence distribution feature vector. In this way, the business occurrence distribution feature vector (may (Including multi-dimensional feature vectors) to predict the business volume in a predetermined period of time, and the final forecast result can be adjusted accordingly when the short-term business volume fluctuates greatly, effectively preventing the business volume from affecting the final short-term fluctuation. The situation of the prediction result occurs, and by using the feature vector of the business volume distribution, the change trend of the business volume (such as the trend of the business volume in the historical business data) can be better captured. This can improve the forecast of business volume. Measurement accuracy, reducing business risks and improving capital utilization efficiency.
The specific embodiments of the present specification have been described above. Other embodiments are within the scope of the appended patent applications. In some cases, the actions or steps described in the scope of the patent application may be performed in a different order than in the embodiments and still achieve the desired result. In addition, the processes depicted in the figures do not necessarily require the particular order shown or sequential order to achieve the desired result. In certain embodiments, multi-tasking and parallel processing are also possible or may be advantageous.
In the 1990s, for a technical improvement, it can be clearly distinguished whether it is an improvement in hardware (for example, the improvement of circuit structures such as diodes, transistors, switches, etc.) or an improvement in software (for method and process Improve). However, with the development of technology, the improvement of many methods and processes can be regarded as a direct improvement of the hardware circuit structure. Designers almost always get the corresponding hardware circuit structure by programming the improved method flow into the hardware circuit. Therefore, it cannot be said that the improvement of a method flow cannot be realized by a hardware entity module. For example, a programmable logic device (Programmable Logic Device (PLD)) (such as a Field Programmable Gate Array (FPGA)) is such an integrated circuit whose logic function is determined by the user's programming of the device. Designers program themselves to "integrate" a digital system on a PLD, without having to ask a chip manufacturer to design and fabricate a dedicated integrated circuit chip. Moreover, today, instead of making integrated circuit chips manually, this programming is mostly implemented with "logic compiler" software, which is similar to the software compiler used in program development and writing. The source code before compilation must also be written in a specific programming language. This is called the Hardware Description Language (HDL), and there is not only one kind of HDL, but many types, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc. Currently, the most commonly used are Very-High-Speed Integrated Circuit Hardware Description Language (VHDL) and Verilog. Those skilled in the art should also be clear that as long as the method flow is logically programmed and integrated into the integrated circuit using the above-mentioned several hardware description languages, the hardware circuit that implements the logic method flow can be easily obtained.
The controller may be implemented in any suitable manner, for example, the controller may take the form of a microprocessor or processor and a computer-readable storage of computer-readable code (such as software or firmware) executable by the (micro) processor. Media, logic gates, switches, application specific integrated circuits (ASICs), programmable logic controllers and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20 and Silicone Labs C8051F320, the memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art also know that, in addition to implementing the controller in a pure computer-readable code manner, the controller can be controlled by logic gates, switches, dedicated integrated circuits, and programmable logic by programming the method steps logically. Controller and embedded microcontroller to achieve the same function. Therefore, the controller can be considered as a hardware component, and the device included in the controller for implementing various functions can also be considered as a structure in the hardware component. Or even, a device for implementing various functions can be regarded as a structure that can be both a software module implementing the method and a hardware component.
The system, device, module, or unit described in the foregoing embodiments may be specifically implemented by a computer chip or entity, or by a product having a certain function. A typical implementation is a computer. Specifically, the computer may be, for example, a personal computer, a laptop, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or A combination of any of these devices.
For the convenience of description, when describing the above device, the functions are divided into various units and described separately. Of course, when implementing one or more embodiments of the present specification, the functions of each unit may be implemented in the same software or hardware.
Those skilled in the art should understand that the embodiments of the present specification may be provided as a method, a system, or a computer program product. Therefore, one or more embodiments of the present specification may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, one or more embodiments of this specification may be implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) containing computer-usable code therein. In the form of a computer program product.
Embodiments of the present specification are described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present specification. It should be understood that each flow and / or block in the flowchart and / or block diagram, and a combination of the flow and / or block in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to generate a machine for instructions executed by the processor of the computer or other programmable data processing device Generate means for implementing the functions specified in one or more flowcharts and / or one or more blocks of the block diagram.
These computer program instructions may also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including a command device , The instruction device implements the functions specified in a flowchart or a plurality of processes and / or a block or a block of the block diagram.
These computer program instructions can also be loaded on a computer or other programmable data processing equipment, so that a series of operating steps can be performed on the computer or other programmable equipment to generate computer-implemented processing, so that the computer or other programmable equipment can The instructions executed on the steps provide steps for implementing the functions specified in one or more flowcharts and / or one or more blocks of the block diagram.
In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.
Memory may include non-persistent memory, random access memory (RAM), and / or non-volatile memory in computer-readable media, such as read-only memory (ROM) or flash memory (flash) RAM). Memory is an example of a computer-readable medium.
Computer-readable media includes permanent and non-permanent, removable and non-removable media. Information can be stored by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), and other types of random access memory (RAM) , Read-only memory (ROM), electrically erasable and programmable read-only memory (EEPROM), flash memory or other memory technology, read-only disc read-only memory (CD-ROM), digital multifunction Optical discs (DVDs) or other optical storage, magnetic tape cartridges, magnetic tape storage or other magnetic storage devices or any other non-transmitting media may be used to store information that can be accessed by computing devices. According to the definition in this article, computer-readable media does not include temporary computer-readable media (transitory media), such as modulated data signals and carrier waves.
It should also be noted that the terms "including,""including," or any other variation thereof are intended to encompass non-exclusive inclusion, so that a process, method, product, or device that includes a range of elements includes not only those elements, but also Other elements not explicitly listed, or those that are inherent to such a process, method, product, or device. Without more restrictions, the elements defined by the sentence "including a ..." do not exclude the existence of other identical elements in the process, method, product or equipment including the elements.
Those skilled in the art should understand that the embodiments of the present specification may be provided as a method, a system, or a computer program product. Therefore, one or more embodiments of the present specification may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, one or more embodiments of this specification may be implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) containing computer-usable code therein. In the form of a computer program product.
One or more embodiments of the present specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. One or more embodiments of the present specification may also be practiced in distributed computing environments in which tasks are performed by remote processing devices connected through a communication network. In a decentralized computing environment, program modules can be located in local and remote computer storage media, including storage devices.
Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple. For the relevant part, refer to the description of the method embodiment.
The above are only examples of the present specification, and are not intended to limit the application. For those skilled in the art, this application may have various modifications and changes. Any modification, equivalent replacement, and improvement made within the spirit and principle of this application shall be included in the scope of the patent application for this application.

S102, S104, S106‧‧‧ steps

S302, S304, S306, S308, S310, S312, S314, S316‧‧‧ steps

401‧‧‧Processing Module

402‧‧‧Business Volume Forecast Module

501‧‧‧ processor

502‧‧‧Memory

503‧‧‧ Power

504‧‧‧Wired or wireless network interface

505‧‧‧I / O interface

506‧‧‧Keyboard

In order to more clearly explain the embodiments of the present specification or the technical solutions in the prior art, the drawings used in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only For some ordinary people skilled in the art, some embodiments described in the specification can also obtain other drawings according to these drawings without paying creative labor.

FIG. 1 is an embodiment of a method for predicting the amount of service occurrence in this specification;

FIG. 2 is a schematic structural diagram of a system for predicting the amount of traffic generated in this specification;

FIG. 3 is another embodiment of a method for predicting the amount of business generated in this specification;

FIG. 4 is an embodiment of a device for predicting the amount of traffic generated in this specification;

FIG. 5 is an embodiment of a device for predicting the amount of traffic generated in this specification.

Claims (13)

A method for predicting the occurrence of business, the method includes: Discretizing historical business data before a predetermined period of time to obtain a time-granularity business volume vector, and generating a business volume distribution feature vector according to continuous-time business volume in the historical business data; Determining the amount of business occurrences within the predetermined time period according to the time-granularity business occurrence vector and the business occurrence distribution feature vector. The method according to item 1 of the scope of patent application, wherein the historical business data includes a first historical business data of a first time period closest to the predetermined time period and a first historical business data other than the first historical business data. Historical business information, The generating a feature vector of a service occurrence distribution according to the service occurrence vector of the time granularity and the continuous occurrence of the service occurrence in the historical service data includes: Generating a first feature vector of the distribution of service occurrences according to the continuous service occurrences in the first historical service data; And generating a second feature vector of the distribution of business occurrences according to the continuous business occurrences in the second historical business data. The method according to item 2 of the scope of patent application, wherein the service occurrence vector and the service occurrence distribution feature vector according to the time granularity are determined before the service occurrence amount in the predetermined time period is determined. The method also includes: Determining a similarity between the first feature vector and the second feature vector; Determine the weight of the second feature vector according to the similarity between the first feature vector and the second feature vector. The method according to item 3 of the scope of patent application, wherein determining the similarity between the first feature vector and the second feature vector includes: The similarity between the first feature vector and the second feature vector is determined by any of the following methods: Euclidean distance, the cosine of the angle of the vector, and the absolute value of the difference of the vector. The method according to item 3 of the scope of patent application, wherein the determining the service occurrence amount in the predetermined time period according to the service occurrence amount vector and the service occurrence distribution feature vector according to the time granularity includes: Merging the time-granularity business volume vector with the first feature vector and the second feature vector to obtain a combined first feature vector and a second feature vector; Based on the second feature vector and the weight of the second feature vector, optimizing initial parameters through a loss function and a predetermined parameter optimization algorithm to obtain optimized initial parameters; Determine the amount of traffic generated in the predetermined time period according to the optimized initial parameters and the first feature vector. The method according to item 5 of the scope of patent application, wherein the predetermined parameter optimization algorithm includes a gradient descent algorithm, Newton method, quasi-Newton method, conjugate gradient method, and heuristic optimization algorithm. A device for predicting the occurrence of traffic, the device includes: A processing module, configured to discretize historical business data before a predetermined period of time, to obtain a time granularity business volume vector, and to generate a business volume distribution characteristic based on the continuous business volume in the historical business data vector; The traffic occurrence forecasting module is configured to determine a traffic occurrence in the predetermined time period based on the time granularity traffic occurrence vector and the traffic occurrence distribution feature vector. The device according to item 7 of the scope of patent application, wherein the historical service data includes a first historical service data in a first time period closest to the predetermined time period and a first Historical business information, The processing module includes: A first feature vector generating unit, configured to generate a first feature vector of a service volume distribution according to a service volume of continuous time in the first historical service data; A second feature vector generating unit is configured to generate a second feature vector of a service volume distribution according to the service volume of continuous time in the second historical service data. The device according to item 8 of the scope of patent application, wherein the device further comprises: A similarity determination module, configured to determine a similarity between the first feature vector and the second feature vector; A weight determining module is configured to determine a weight of the second feature vector according to a similarity between the first feature vector and the second feature vector. The device according to item 9 of the scope of patent application, wherein the similarity determination module is configured to determine the similarity between the first feature vector and the second feature vector through any one of the following devices: European The absolute value of the distance, the angle cosine of the vector, and the difference of the vector. The device according to item 9 of the scope of patent application, wherein the business occurrence prediction module includes: A merging unit for merging the time-granularity business volume vector with the first feature vector and the second feature vector to obtain a merged first feature vector and a second feature vector; An initial parameter optimization unit, configured to optimize the initial parameters based on the second feature vector through a loss function and a predetermined parameter optimization algorithm to obtain optimized initial parameters; A traffic occurrence prediction unit is configured to determine a traffic occurrence in the predetermined time period according to the optimized initial parameters and the first feature vector. The device according to item 11 of the scope of patent application, wherein the predetermined parameter optimization algorithm includes a gradient descent algorithm, Newton method, quasi-Newton method, conjugate gradient method, and heuristic optimization algorithm. A device for predicting the amount of business occurrences. The device for predicting the amount of business occurrences includes: Processor; and Memory arranged to store computer-executable instructions that, when executed, cause the processor to: Discretize historical business data before a predetermined period of time to obtain a time granularity business volume vector, and generate a business volume distribution feature vector based on the continuous business volume in the historical business data; Determining the amount of business occurrences in the predetermined time period according to the time-granularity business occurrence vector and the business occurrence distribution feature vector.