A service model optimization method based on confidence perception
Through the service model optimization method based on confidence perception, dynamically adjusting the search speed and stride, optimizing business logic and service distribution, the problem of ignoring platform influence and insufficient collaborative analysis in service model optimization in the existing technology is solved, and more efficient Service model optimization effect.
Patent Information
- Application Number
- CN202210401977.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-04-18
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2042-04-18
AI Technical Summary
The existing technology ignores the impact of the service deployment platform on the model performance in service model optimization, and lacks a comprehensive analysis of participant collaboration, resulting in poor service quality and resource, data, and value transmission efficiency.
The service model optimization method based on confidence perception is adopted, and the target mode is searched cyclically by initializing the candidate list, temperature and confidence values, dynamically adjusting the search speed and stride, and optimizing the business logic and service distribution to improve the overall performance of the service model.
Significantly save optimization time, improve optimization effect, take into account time, cost, model entropy and data, resource and value transmission efficiency, and achieve better overall service model orchestration.
Smart Images

Figure CN114818299B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of service mode calculation in software engineering, and in particular relates to a service mode optimization method based on confidence perception. Background Art
[0002] Compared with traditional web services, current complex service systems are usually composed of autonomous entities (called participants) from different fields and cloud services on different servers. In these complex service systems, participants can refer to virtual users or intelligent agents that need to collaborate and interact with other entities to continue the service process. In the collaborative process, data, resources and value are transferred and exchanged between participants, and the overall paradigm used to describe this process is called the service model.
[0003] Service patterns refer to the collaboration between different participants and the exchange of data, resources and value in complex service systems. Service patterns originate from the extension of workflows and business processes. To support the analysis of complex service systems, service patterns complement traditional service modeling methods by describing the business from four aspects: workflow, data flow, resource flow and value flow.
[0004] For example, a Chinese patent document with publication number CN111612330A discloses a quantitative evaluation method for service models for cross-border services: define the top-level elements of the service model, including defining participants, workflows, data flows, resource flows, and cash flows; describe the participants in the service model; describe the workflows between participants based on the existing participants; describe the data flows between participants based on the workflows between participants; describe the resource flows between participants based on the workflows between participants; describe the cash flows between participants based on the workflows between participants; calculate the evaluation indicators of the service model based on the described workflows, data flows, resource flows, and cash flows, including running time, consumption cost, delivery efficiency, value, and reliability; and calculate the model entropy based on the evaluation indicators for overall evaluation of the service model. This method can help product managers, entrepreneurs, business consultants, and business designers to quantitatively evaluate existing service models.
[0005] An important focus of service model analysis is service model optimization, which plays a vital role in the iteration of complex service systems. Previous studies have done a lot of work on service process optimization, but due to the limitations of the service model description ability, they have ignored the impact of the service deployment platform on the performance of the service model.
[0006] In addition, there are studies that consider resource capacity and infrastructure cost and propose the Cloud Service Distribution Problem (CSDP) to optimize cloud service coordination in cloud network environments. However, a comprehensive analysis of the collaboration between participants is still missing.
[0007] In order to solve these problems, the goal of service model optimization is not only to improve the quality of service (QoS), but also to focus on the efficiency of data, resource and value transfer and the rationality of the overall orchestration of the service model. Therefore, it is urgent to design a new service model optimization method to solve the above problems existing in the existing technology. Summary of the invention
[0008] The present invention provides a service mode optimization method based on confidence perception, which can greatly save optimization time and improve optimization effect.
[0009] A service mode optimization method based on confidence perception includes the following steps:
[0010] (1) Input the original model Pa to be optimized, where the original model Pa consists of multiple participants and the workflows, data flows, resource flows, and value flows between them;
[0011] (2) Initialize the candidate list PaList of the original pattern Pa;
[0012] (3) Initialization temperature T;
[0013] (4) Initialize confidence C;
[0014] (5) Initialize the maximum number of iterations IterMax;
[0015] (6) Initialize the termination threshold Th;
[0016] (7) According to the pattern optimization index, the target pattern Pa* is searched cyclically, and the number of cycles is IterMax;
[0017] (8) Lowering the temperature T;
[0018] (9) If the pattern Pa* obtained at the end of the loop of step (7) remains consistent for Th consecutive times, then Pa* is the optimized target pattern; otherwise, jump to step (7).
[0019] In step (2), the candidate list PaList is initialized by four copies of Pa.
[0020] In step (3), the calculation formula of the initialization temperature T is: T = K*X 2 ;
[0021] Among them, K is set to a real number between 5 and 10, X 2 It is the variance of the optimization index sequence composed of the original pattern Pa and the pattern generated after random search.
[0022] In step (4), confidence C is an arbitrary real number. When it is a negative number, it indicates that the optimization potential of the current evolution direction is weak. When C is a positive number, it indicates that the optimization potential of the current evolution direction is strong. The initialization value of confidence C is 0.
[0023] In step (5), the formula for initializing the maximum number of iterations IterMax is:
[0024] IterMax=(|Activity| / (|Event|+|Gateway|))! *(|Event|+|Gateway|)
[0025] Among them, |Activity|, |Event|, and |Gateway| are the number of activities, events, and gateways contained in the optimized pattern, respectively, and ! represents factorial.
[0026] In step (6), the initialization value of the termination threshold Th is rounded up to the integer of log(IterMax).
[0027] In step (7), the model optimization index includes 6 individual indexes and 1 overall index; the 6 individual indexes are model operation time Ti, model operation cost Co, model entropy En, data transfer efficiency DaEf, resource transfer efficiency ReEf and value transfer efficiency VaEf; the overall index is model loss, and the formula is:
[0028] Lo=(log(Ti)+log(Co))×3En / (DaEf+ReEf+VaEf)
[0029] Among them, the smaller the model operation time Ti, the model operation cost Co, the model entropy En and the model loss Lo, the better; the larger the data transmission efficiency DaEf, the resource transmission efficiency ReEf and the value transmission efficiency VaEf, the better.
[0030] In step (7), when the target pattern Pa* is searched cyclically, the steps of each search are:
[0031] (7-1) Set the search step St. When C is greater than or equal to 0, St is 1. When C is less than 0, St is the minimum value between 1-C and |Activity| / 2.
[0032] (7-2) For the optimization pattern Pa, the execution order of activities is exchanged St times, and Pa is obtained. f ;
[0033] (7-3) For the optimization mode Pa, St activity execution platforms are exchanged to obtain Pa d ;
[0034] (7-4) Simultaneously perform the transformations in steps (7-2) and (7-3) on the optimization pattern Pa to obtain Pa h ;
[0035] (7-5) Let pattern candidate list PaList = [Pa, Pa f ,Pa d ,Pa h ];
[0036] (7-6) Compare the single indexes of the model optimization in the model candidate list in the previous iteration and the model candidate list in this iteration one by one; if the model index in this iteration is better, add 1 to the confidence C; if the model index in this iteration is worse, subtract 1 from the confidence C;
[0037] (7-7) If the value of confidence C remains unchanged or greater after step (7-6), the pattern candidate list generated in this round is used, otherwise the pattern candidate list generated in the previous iteration is used as PaList;
[0038] (7-8) Compare the optimization indicators of the four modes in PaList. If the mode with the best optimization indicator is Pa f 、Pa d 、Pa h If one of the modes is found, the mode Pa to be optimized is adopted to enter the next round of iteration; if the mode with the best optimization index is Pa, then according to e DIFF / T The probability of using Pa f 、Pa d 、Pa h The optimal mode in is the mode to be optimized Pa and enters the next round of iteration;
[0039] Among them, DIFF = -|Pa optimization index value - Pa f ,Pa d ,Pa h The average value of the optimization index |, otherwise Pa remains in the optimization mode.
[0040] In step (8), the temperature T is reduced by:
[0041] T = T / log (1 + IterT + reg (C))
[0042] Where IterT is the number of times the temperature T has dropped, reg(C) = (e C -1) / (e C +1).
[0043] Compared with the prior art, the present invention has the following beneficial effects:
[0044] 1. The present invention can optimize the service model as a whole through business logic optimization and service distribution optimization, which improves the final optimization effect compared with the traditional method of only optimizing business logic or service distribution.
[0045] 2. The present invention can optimize the time, cost, and pattern entropy of the pattern while taking into account the transmission efficiency of data, resources, and value in the service pattern through the overall indicator pattern loss.
[0046] 3. By introducing a confidence mechanism, the present invention can dynamically adjust the search speed and search step in search spaces with different optimization potentials, which greatly saves optimization time and improves the optimization effect. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] Figure 1 A flow chart of a service mode optimization method based on confidence perception according to the present invention;
[0048] Figure 2 This is a schematic diagram of an initial service mode introduced in an embodiment of the present invention;
[0049] Figure 3 Schematic diagram of a service mode optimized by the optimization method of the present invention in an embodiment of the present invention. DETAILED DESCRIPTION
[0050] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be pointed out that the embodiments described below are intended to facilitate the understanding of the present invention and do not have any limiting effect on the present invention.
[0051] like Figure 1 As shown, a service mode optimization method based on confidence perception includes the following steps:
[0052] Step (1), input the original pattern Pa to be optimized.
[0053] This embodiment takes the online travel reservation mode as an example. Figure 2 As shown in the figure, the customer registers for travel, purchases transportation tickets, and applies for insurance in sequence. Between each operation, the customer needs to pay the financial institution first, and then the travel platform, transportation company, and insurance company confirm the request and transfer the travel voucher, transportation ticket, and insurance certificate. After all these activities, the customer has the opportunity to notify the financial institution to cancel the entire travel application, otherwise the financial institution will settle the value paid by the customer with all stakeholders.
[0054] The disadvantage of the original model is that after almost every activity, the work needs to be handed over to another participant and the next activity needs to be completed on a different server, which results in a lot of extra time and cost.
[0055] Step (2), initialize the candidate list PaList to four copies of the original pattern Pa.
[0056] Step (3), by sampling the optimized overall index Lo of the original mode Pa multiple times, the initialization temperature T = 0.8346.
[0057] Step (4), initialize confidence C=0.
[0058] Step (5), initialize the maximum number of iterations IterMax = (13 / (3+1))! *(3+1) = 33.14, and take 34.
[0059] Step (6), initialize the termination threshold Th = log (34) = 3.5264, and take 4.
[0060] Step (7), according to the pattern optimization index, cyclically search for the target pattern Pa*, this step is cyclically executed IterMax=34 times.
[0061] (7-1) Since C = 0, set the search step St = 1;
[0062] (7-2) The execution order of the activities to be optimized Pa is exchanged St = 1 times to obtain Pa f ;
[0063] (7-3) The optimization pattern Pa is exchanged St = 1 times for the active execution platform, and Pa is obtained d ;
[0064] (7-4) Simultaneously perform the transformations in (7-2) and (7-3) on the optimization pattern Pa to obtain Pa h ;
[0065] (7-5) Let pattern list PaList = [Pa,Pa f ,Pa d ,Pa h ];
[0066] (7-6) For the pattern list [Pa,Pa,Pa,Pa] in the previous iteration and the pattern list [Pa,Pa f ,Pa d ,Pa h ] are compared one by one. For example, in this round, the performance of Pa is {Ti:1502.33, Co:313.44, En:1.4729, DaEf:12.56, ReEf:3.99, VaEf:12.00}, Pa fThe performance is {Ti:1483.60,Co:364.30,En:1.5292,DaEf:13.44,ReEf:2.61,VaEf:11.71},Pa d The performance is {Ti:1295.66,Co:280.15,En:1.4544,DaEf:16.42,ReEf:3.70,VaEf:11.85},Pa h The performance is {Ti:1293.51,Co:262.57,En:1.4977,DaEf:16.68,ReEf:4.90,VaEf:10.94}. First, compare the first item in the list in this round with that in the previous round of iterations. Since they are the same, the confidence C remains unchanged. Then compare the second item in the list in this round with that in the previous round of iterations. Ti and DaEf in this round are better, and the confidence value is increased by 2. Co, En, ReEf, and VaEf in this round are worse, and the confidence C is reduced by 4, so that the confidence C=0+2-4=-2. Compare the third item in the list in this round with that in the previous round of iterations. Ti, Co, En, and DaEf in this round are better, and the confidence value is increased by 4. ReEf and VaEf in this round are worse, and the confidence C is reduced by 2, so that the confidence C=-2+4-2=0. Compare the fourth item in the list in this round with that in the previous round of iterations. Ti, Co, DaEf, and ReEf in this round are better, and the confidence value is increased by 4. En and VaEf in this round are worse, and the confidence C is reduced by 2, so that the confidence C=0+4-2=2.
[0067] (7-7) The confidence C value increases after (7-6), so the pattern list PaList generated in this round is used = [Pa, Pa f ,Pa d ,Pa h ];
[0068] (7-8) Comparing the optimization indexes of the four modes in PaList, the indexes are 2.0212, 2.1816, 1.7480, and 1.7599 respectively. If the mode with the best optimization index is Pa d , adopt this mode as the mode to be optimized Pa to enter the next round of iteration.
[0069] Step (8), since it is the first time to reduce the temperature, IterT = 1, after the cyclic search in step (7), the confidence C = 47, so the temperature after the reduction is T = 0.8346 / (1+1+(e 47 -1) / (e 47 +1))=0.2782.
[0070] Step (9), when the pattern Pa* obtained after the loop of step (7) is completed and remains consistent for Th=4 consecutive times, Pa* is obtained as the optimized target pattern found.
[0071] Figure 3 This is the optimization target model obtained in an experiment. It is obvious that the business logic and service distribution have been optimized. The business logic has become that the customer makes all the orders first, and then the four service providers do the rest, which greatly reduces the cooperation time between the participants. In addition, adjacent activities are also more often carried out on the same server, which reduces costs. And in the process of algorithm optimization, the transmission efficiency of data, resources, and value as well as the overall performance are also taken into account. The final optimization target model is Lo = 1.6296, Ti = 1173.36, Co = 197.69, En = 1.4347, DaEf = 16.83, ReEf = 4.34, VaEf = 11.48. It can be seen that the optimized service model performs better in Lo, Ti, Co, En, DaEf, and ReEf.
[0072] In addition, the method of the present invention is compared with three traditional methods, namely, the existing simulated annealing method, the method of only performing service distribution optimization, and the method of only performing business logic optimization. Each method is optimized 1000 times, and the average value is finally taken for comparison. The results are shown in Table 1.
[0073] Table 1
[0074]
[0075] It can be seen that the method of the present invention performs better than other methods in terms of the comprehensive index Lo, and requires the least number of iterative search rounds on average.
[0076] The embodiments described above provide a detailed description of the technical solutions and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, supplements and equivalent substitutions made within the scope of the principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A service mode optimization method based on confidence perception, characterized in that: The following steps are involved: (1) Input the original model Pa to be optimized, where the original model Pa consists of multiple participants and the workflows, data flows, resource flows, and value flows between them; (2) Initialize the candidate list PaList of the original pattern Pa; (3) Initialization temperature T; (4) Initialize confidence C; (5) Initialize the maximum number of iterations IterMax; (6) Initialize the termination threshold Th; (7) According to the pattern optimization index, the target pattern Pa* is searched cyclically, and the number of cycles is IterMax; the pattern optimization index includes 6 individual indicators and 1 overall indicator; the 6 individual indicators are pattern operation time Ti, pattern operation cost Co, pattern entropy En, data transfer efficiency DaEf, resource transfer efficiency ReEf and value transfer efficiency VaEf; the overall indicator is pattern loss, and the formula is: Lo=(log(Ti)+log(Co))×3En / (DaEf+ReEf+VaEf) Among them, the smaller the mode operation time Ti, the mode operation cost Co, the mode entropy En, and the mode loss Lo, the better; the larger the data transmission efficiency DaEf, the resource transmission efficiency ReEf, and the value transmission efficiency VaEf, the better; When cyclically searching for the target pattern Pa*, the steps of each search are: (7-1) Set the search step St. When C is greater than or equal to 0, St is 1. When C is less than 0, St is the minimum value between 1-C and |Activity| / 2. |Activity| is the number of activities contained in the optimized pattern. (7-2) For the optimization pattern Pa, the execution order of activities is exchanged St times, and Pa is obtained. f ; (7-3) For the optimization mode Pa, St activity execution platforms are exchanged to obtain Pa d ; (7-4) Simultaneously perform the transformations in steps (7-2) and (7-3) on the optimization pattern Pa to obtain Pa h ; (7-5) Let pattern candidate list PaList = [Pa, Pa f ,Pa d ,Pa h ]; (7-6) Compare the single indexes of the model optimization in the model candidate list in the previous iteration and the model candidate list in this iteration one by one; if the model index in this iteration is better, add 1 to the confidence C; if the model index in this iteration is worse, subtract 1 from the confidence C; (7-7) If the value of confidence C remains unchanged or greater after step (7-6), the pattern candidate list generated in this round is used, otherwise the pattern candidate list generated in the previous iteration is used as PaList; (7-8) Compare the optimization indicators of the four modes in PaList. If the mode with the best optimization indicator is Pa f 、Pa d 、Pa h If one of the modes is found, the mode Pa to be optimized is adopted to enter the next round of iteration; if the mode with the best optimization index is Pa, then according to e DIFF / T The probability of using Pa f 、Pa d 、Pa h The optimal mode in is the mode to be optimized Pa and enters the next round of iteration; Among them, DIFF = -|Pa optimization index value - Pa f ,Pa d ,Pa h The average value of the optimization index |, otherwise Pa is kept in the optimization mode; (8) Lowering the temperature T; the method for lowering the temperature T is: T = T / log (1 + IterT + reg (C)) Where IterT is the number of times the temperature T has dropped, reg(C) = (e C -1) / (e C +1); (9) If the pattern Pa* obtained at the end of the loop of step (7) remains consistent for Th consecutive times, then Pa* is the optimized target pattern; otherwise, jump to step (7).
2. The service mode optimization method based on confidence perception according to claim 1 is characterized in that: In step (2), the candidate list PaList is initialized by making four copies of Pa.
3. The service mode optimization method based on confidence perception according to claim 1 is characterized in that: In step (3), the calculation formula of the initialization temperature T is: T = KX 2 ; Among them, K is set to a real number between 5 and 10, X 2 It is the variance of the optimization index sequence composed of the original pattern Pa and the pattern generated after random search.
4. The service mode optimization method based on confidence perception according to claim 1 is characterized in that: In step (4), confidence C is an arbitrary real number. When it is a negative number, it indicates that the optimization potential of the current evolution direction is weak. When C is a positive number, it indicates that the optimization potential of the current evolution direction is strong. The initialization value of confidence C is 0.
5. The service mode optimization method based on confidence perception according to claim 1 is characterized in that: In step (5), the formula for initializing the maximum number of iterations IterMax is: IterMax=(|Activity| / (|Event|+|Gateway|))! ×(|Event|+|Gateway|) Among them, |Activity|, |Event|, and |Gateway| are the number of activities, events, and gateways contained in the optimized pattern, respectively, and ! represents factorial.
6. The service mode optimization method based on confidence perception according to claim 1 is characterized in that: In step (6), the initialization value of the termination threshold Th is rounded up to the integer of log(IterMax).
