Data processing method, device, equipment and medium based on multi-model computing fusion
By acquiring and fusing the results of multiple computing models, dynamically adjusting the algorithm model according to the prediction period and item characteristics, the problem of resource waste and model fusion strategies in the existing technology is solved, and the accuracy of sales forecasts in the retail supply chain is improved. efficiency.
Patent Information
- Application Number
- CN202010580968.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2020-06-23
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2040-06-23
AI Technical Summary
In retail supply chain scenarios, the existing technology cannot refinely control the algorithm model of each item, resulting in waste of computing resources and model fusion strategies that cannot be dynamically adjusted, and personalized processing cannot be performed based on item characteristics and predicted periods.
By obtaining the model configuration table of items to be processed, determining the target calculation model that matches the predicted period, and fusing multiple calculation results according to the fusion priority and category, the calculation model is dynamically adjusted to improve accuracy and resource utilization.
It reduces the resources and duration of algorithm execution, improves the accuracy of calculation results, and facilitates personalized adaptation of items and control of model fusion categories, achieving a more fine-grained prediction process.
Smart Images

Figure CN112288457B_ABST
Abstract
Description
Background Art
[0002] In retail supply chain scenarios, for items for sale, it is necessary to predict the sales volume of each item in the future every day, and then make reasonable replenishment based on the predicted sales volume to reduce inventory turnover. In some large retail scenarios, the items sold cover almost all categories of items. However, the characteristics of each item, marketing strategy, external influences, etc. lead to different sales results in different periods. No algorithm model can adapt to all items and the differences of each item in different periods.
[0003] In the prior art, one or more algorithm models are usually used to predict the sales volume of all items in the future, and then the prediction results of multiple models are integrated to obtain the final sales volume prediction result. Summary of the invention
[0004] The purpose of the present disclosure is to provide a data processing method based on multi-model computational fusion, a data processing device based on multi-model computational fusion, an electronic device and a computer-readable storage medium, thereby at least to a certain extent overcoming the problem that when using multiple algorithm models to calculate the processed items, the algorithm model adapted for each item cannot be finely controlled according to the prediction period or other item information, resulting in wasted computing resources and inability to dynamically adjust the fusion category of the model.
[0005] Other features and advantages of the present disclosure will become apparent from the following detailed description, or may be learned in part by the practice of the present invention.
[0006] According to a first aspect of the present disclosure, there is provided a data processing method based on multi-model computational fusion, comprising: obtaining a model configuration table of an object to be processed; wherein the model configuration table comprises at least one candidate computational model corresponding to the object to be processed; determining a time period to be predicted for the object to be processed, and determining at least one target computational model that matches the object to be processed from at least one candidate computational model according to the time period to be predicted; if there are multiple target computational models, inputting the execution data of the object to be processed into each target computational model respectively, and each target computational model outputs the computational result of the object to be processed; respectively determining the fusion priority and fusion category of each target computational model, so as to fuse the multiple computational results according to each fusion priority and each fusion category, and obtain the prediction result of the object to be processed.
[0007] Optionally, before obtaining the model configuration table of the object to be processed, the above method also includes: obtaining item information of the object to be processed; determining at least one candidate calculation model based on the item information, and determining model information of at least one candidate calculation model; generating a model configuration table of the object to be processed based on the determined model information and item information.
[0008] Optionally, at least one target computing model that matches the object to be processed is determined from at least one candidate computing model according to the time period to be predicted, including: performing the following operation steps on the candidate computing models one by one: obtaining and determining the verification mode of the candidate computing model; if the verification mode is a fixed verification mode, obtaining a preconfigured computing model corresponding to the object to be processed, and using the preconfigured computing model as the target computing model; if the verification mode is an automatic verification mode, determining whether the candidate computing model matches the object to be processed according to the time period to be predicted and the object information of the object to be processed; if the candidate computing model matches the object to be processed, determining the candidate computing model as the target computing model.
[0009] Optionally, multiple calculation results are fused according to each fusion priority and each fusion category to obtain a prediction result for the item to be processed, including: dividing the time period to be predicted into multiple unit calculation days, and determining the unit calculation results corresponding to each unit calculation day; fusing each unit calculation result according to each fusion priority and each fusion category to obtain a unit prediction result; and taking all unit prediction results in the time period to be predicted as the prediction result.
[0010] Optionally, each unit calculation result is fused according to each fusion priority and each fusion category to obtain a unit prediction result, including: sorting each fusion priority to obtain a level sorting result; determining the highest priority from the level sorting result; wherein the highest priority is the fusion priority with the largest priority value; if the number of highest priorities is one, the unit calculation result of the target calculation model with the highest priority is used as the unit prediction result.
[0011] Optionally, each unit calculation result is fused according to each fusion priority and each fusion category to obtain a unit prediction result, and also includes: if there is a multiple number of highest priorities, the unit calculation results of the target calculation models corresponding to the multiple highest priorities are used as the results to be fused; and each result to be fused is fused according to the fusion category to obtain a unit prediction result.
[0012] Optionally, each result to be fused is fused according to the fusion category to obtain a unit prediction result, including: if the fusion category is mean fusion, then multiple results to be fused are mean processed to obtain a unit prediction result; if the fusion category is proportional fusion, then the calculation weights of each result to be fused are determined, and the unit prediction result is obtained according to each calculation weight and each result to be fused; wherein; determining the calculation weight of each result to be fused includes: determining the calculation error of each target calculation model, and determining the calculation weight of each result to be fused according to each calculation error.
[0013] According to a second aspect of the present disclosure, a data processing device based on multi-model computational fusion is provided, comprising: an information acquisition module, used to acquire a model configuration table of an item to be processed; wherein the model configuration table includes at least one candidate computational model corresponding to the item to be processed; a model determination module, used to determine a time period to be predicted for the item to be processed, and determine at least one target computational model matching the item to be processed from at least one candidate computational model according to the time period to be predicted; a computational module, used to input the execution data of the item to be processed into each target computational model respectively if there are multiple target computational models, and each target computational model outputs the computational result of the item to be processed; a result determination module, used to respectively determine the fusion priority and fusion category of each target computational model, so as to fuse the multiple computational results according to each fusion priority and each fusion category to obtain a prediction result of the item to be processed.
[0014] Optionally, the data processing device based on multi-model computing fusion also includes a model configuration table generation module, which is used to obtain item information of the item to be processed; determine at least one candidate computing model based on the item information, and determine the model information of at least one candidate computing model; generate a model configuration table of the item to be processed based on the determined model information and item information.
[0015] Optionally, the model determination module includes a model determination unit, which is used to obtain and determine the verification mode of the candidate calculation model; if the verification mode is a fixed verification mode, then the preconfigured calculation model corresponding to the item to be processed is obtained, and the preconfigured calculation model is used as the target calculation model; if the verification mode is an automatic verification mode, then it is determined whether the candidate calculation model matches the item to be processed based on the predicted time period and the item information of the item to be processed; if the candidate calculation model matches the item to be processed, the candidate calculation model is determined as the target calculation model.
[0016] Optionally, the result determination module includes a result determination unit, which is used to divide the time period to be predicted into multiple unit calculation days, determine the unit calculation results corresponding to each unit calculation day; fuse each unit calculation result according to each fusion priority and each fusion category to obtain a unit prediction result; and take all unit prediction results in the time period to be predicted as the prediction result.
[0017] Optionally, the result determination unit includes a first result determination sub-unit, which is used to sort each fusion priority to obtain a level sorting result; determine the highest priority from the level sorting result; wherein the highest priority is the fusion priority with the largest priority value; if the number of highest priorities is one, the unit calculation result of the target calculation model with the highest priority is used as the unit prediction result.
[0018] Optionally, the result determination unit also includes a second result determination sub-unit, which is used to, if there is a plurality of highest priority levels, use the unit calculation results of the target calculation models corresponding to the plurality of highest priority levels as the results to be fused; and fuse each of the results to be fused according to the fusion category to obtain a unit prediction result.
[0019] Optionally, the result determination unit includes a result determination sub-unit, which is used to perform mean processing on multiple results to be fused to obtain a unit prediction result if the fusion category is mean fusion; if the fusion category is proportional fusion, determine the calculation weights of each result to be fused, and obtain the unit prediction result according to each calculation weight and each result to be fused; wherein; determining the calculation weight of each result to be fused includes: determining the calculation error of each target calculation model, and determining the calculation weight of each result to be fused according to each calculation error.
[0020] According to a third aspect of the present disclosure, an electronic device is provided, comprising: a processor; and a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, a data processing method based on multi-model computational fusion as described above is implemented.
[0021] According to a fourth aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the data processing method based on multi-model computational fusion as described in any one of the above items is implemented.
[0022] The technical solution provided by the present disclosure may have the following beneficial effects:
[0023] The data processing method based on multi-model computing fusion in the exemplary embodiment of the present disclosure obtains a model configuration table of the items to be processed; wherein the model configuration table includes at least one candidate computing model corresponding to the items to be processed; determines the time period to be predicted for the items to be processed, and determines at least one target computing model that matches the items to be processed from at least one candidate computing model according to the time period to be predicted; if there are multiple target computing models, the execution data of the items to be processed are respectively input into each target computing model, and each target computing model outputs the calculation result of the items to be processed; respectively determines the fusion priority and fusion category of each target computing model, so as to fuse the multiple calculation results according to each fusion priority and each fusion category, and obtain the prediction result of the items to be processed. Through the data processing method based on multi-model computing fusion disclosed in the present disclosure, on the one hand, when using at least one candidate computing model to calculate the items to be processed, the target computing model that matches the time period to be predicted can be determined from at least one candidate computing model, and the resources and time consumed by each algorithm execution can be greatly reduced by calculating the items to be processed with the determined target computing model. On the other hand, when using multiple target calculation models to process the items to be processed, the corresponding calculation model can be dynamically adjusted according to finer-grained constraints such as the time period to be predicted to improve the accuracy of the calculation results. On the other hand, decoupling the personalized adaptation calculation model and fusion category of the item from the standard prediction process can be easily expanded and fine-grained.
[0024] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS
[0025] The drawings herein are incorporated into the specification and constitute a part of the specification, showing embodiments consistent with the present disclosure, and together with the specification, are used to explain the principles of the present disclosure. Obviously, the drawings described below are only some embodiments of the present disclosure, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:
[0026] Figure 1 A flowchart of a data processing method based on multi-model computational fusion according to an exemplary embodiment of the present disclosure is schematically shown;
[0027] Figure 2 Schematically shows an overall structural diagram of processing an article to be processed according to an exemplary embodiment of the present disclosure;
[0028] Figure 3 A flowchart for generating a model configuration table of an object to be processed according to an exemplary embodiment of the present disclosure is schematically shown;
[0029] Figure 4 A flowchart for determining a target computing model corresponding to an object to be processed according to an exemplary embodiment of the present disclosure is schematically shown;
[0030] Figure 5 A flowchart for determining a prediction result of an item to be processed according to some exemplary embodiments of the present disclosure is schematically shown;
[0031] Figure 6 A flowchart for determining a prediction result of an item to be processed according to another exemplary embodiment of the present disclosure is schematically shown;
[0032] Figure 7 A flowchart for determining a prediction result of an item to be processed according to a fusion category according to an exemplary embodiment of the present disclosure is schematically shown;
[0033] Figure 8 A block diagram of a data processing device based on multi-model computing fusion according to an exemplary embodiment of the present disclosure is schematically shown;
[0034] Fig. 9 A block diagram schematically shows an electronic device according to an exemplary embodiment of the present disclosure;
[0035] Fig.10 A schematic diagram of a computer-readable storage medium according to an exemplary embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION
[0036] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that this disclosure will be comprehensive and complete and will fully convey the concepts of the example embodiments to those skilled in the art. The same reference numerals in the figures represent the same or similar parts, and thus their repeated description will be omitted.
[0037] In addition, the described features, structures or characteristics may be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure may be practiced without one or more of the specific details, or other methods, components, devices, steps, etc. may be adopted. In other cases, known structures, methods, devices, implementations, materials or operations are not shown or described in detail to avoid blurring the various aspects of the present disclosure.
[0038] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities may be implemented in software form, or these functional entities or parts of functional entities may be implemented in one or more software hardened modules, or these functional entities may be implemented in different networks and / or processor devices and / or microcontroller devices.
[0039] In the prior art, one or more algorithm models are usually used to predict the sales volume of all items in the future, and then the prediction results of multiple models are integrated to obtain the final sales prediction results. However, when using multiple algorithm models to calculate the items to be processed, it is impossible to finely control the algorithm adapted for each item. Each algorithm model needs to perform prediction calculations on all items, resulting in each algorithm model consuming a lot of resources and a long execution time. Secondly, when fusing the prediction results of multiple models, they can only be integrated according to the model effect, and the model fusion strategy cannot be dynamically and finely controlled according to the characteristics of the item, the prediction time period, etc. Thirdly, when a new algorithm model is used for step-by-step cutting, it is not very convenient to finely and dynamically control the cutting range.
[0040] Based on this, in this example embodiment, a data processing method based on multi-model computing fusion is first provided. The data processing method based on multi-model computing fusion of the present disclosure can be implemented by a server, and the method described in the present disclosure can also be implemented by a terminal device, wherein the terminals described in the present disclosure may include mobile terminals such as mobile phones, tablet computers, laptops, PDAs, personal digital assistants (PDAs), and fixed terminals such as desktop computers. Figure 1 The following is a schematic diagram of a data processing method based on multi-model computing fusion according to some embodiments of the present disclosure. Figure 1 The data processing method based on multi-model computing fusion may include the following steps:
[0041] Step S110, obtaining a model configuration table of the object to be processed; wherein the model configuration table includes at least one candidate calculation model corresponding to the object to be processed.
[0042] Step S120, determining a time period to be predicted for the object to be processed, and determining at least one target computing model matching the object to be processed from at least one candidate computing model according to the time period to be predicted.
[0043] Step S130: If there are multiple target calculation models, the execution data of the object to be processed is input into each target calculation model respectively, and each target calculation model outputs the calculation result of the object to be processed.
[0044] Step S140, respectively determining the fusion priority and fusion category of each target calculation model, so as to fuse multiple calculation results according to each fusion priority and each fusion category to obtain a prediction result of the item to be processed.
[0045] According to the data processing method based on multi-model computing fusion in this example embodiment, on the one hand, when using at least one candidate computing model to calculate the items to be processed, a target computing model that matches the time period to be predicted can be determined from at least one candidate computing model, and the resources and time consumed by each algorithm execution can be greatly reduced by using the determined target computing model to perform computing processing on the items to be processed. On the other hand, when using multiple target computing models to process the items to be processed, the corresponding computing models can be dynamically adjusted according to finer-grained constraints such as the time period to be predicted, thereby improving the accuracy of the calculation results. On the other hand, decoupling the personalized adaptation computing model and fusion category of the items from the standard prediction process can be easily expanded and fine-grained controlled.
[0046] Next, the data processing method based on multi-model computational fusion in this example embodiment will be further described.
[0047] In step S110, a model configuration table of the object to be processed is obtained; wherein the model configuration table includes at least one candidate calculation model corresponding to the object to be processed.
[0048] In some exemplary embodiments of the present disclosure, the item to be processed may be an item for which a computing model is used to perform predictive calculations on relevant feature data of the item. For example, in a vending scenario, the item to be processed may be an item on sale. The model configuration table may be a configuration table composed of relevant information of candidate computing models corresponding to the item to be processed, and the model configuration table may include candidate computing models that can be used to process the item to be processed, and relevant model information of these candidate computing models. The candidate computing model may be a computing model that can be used to process the item to be processed, and the candidate computing model may be a computing model determined based on relevant information of the item to be processed, and the candidate computing model may be used to perform processing operations such as predictive calculations on the item to be processed.
[0049] When processing an item to be processed, you can first obtain a model configuration table corresponding to the item to be processed. The model configuration table may include all candidate computing models that can process the item to be processed. Figure 2 , Figure 2 The overall structure diagram of processing the to-be-processed object according to the exemplary embodiment of the present disclosure is schematically shown. Before obtaining the model configuration table of the to-be-processed object, the model configuration table can be configured according to the object information of the to-be-processed object by the configuration module 210 .
[0050] According to some exemplary embodiments of the present disclosure, item information of an item to be processed is obtained; at least one candidate computing model is determined according to the item information, and model information of at least one candidate computing model is determined; and a model configuration table of the item to be processed is generated according to the determined model information and item information. The item information may be relevant information of the item to be processed, and the item information may include item attributes, relevant sales data, configuration requirements, etc.; specifically, the item attributes may be attributes related to the item to be processed itself, for example, the item attributes may include the name of the item, the genus category, the item number, etc. The relevant sales data may be historical sales data, promotion strategy, etc. of the item to be processed within a certain period of time. The configuration requirements may include configuration granularity, and the configuration granularity may be the granularity used when configuring the model configuration table of the item to be processed. For example, the configuration granularity may include item granularity, and may also include type granularity of the item type corresponding to the item to be processed. The item type, i.e., the item category, may be the genus category corresponding to the item to be processed. For example, if the item to be processed is a "computer", the item type corresponding to the item to be processed is an electronic product. The model information may be model attribute information corresponding to the candidate computing model, for example, the model information may include the verification mode, fusion priority, and fusion category of the candidate computing model, etc. Specifically, the verification mode may indicate whether the candidate calculation model can be dynamically verified to see whether it matches the item to be processed according to the predicted time period of the item to be processed, and the verification mode may include an automatic verification mode and a fixed verification mode. The fusion priority may be the priority used when determining whether to perform fusion processing on the calculation result after the candidate calculation model obtains the calculation result. The fusion category may be the category corresponding to the fusion category of the calculation result calculated by the candidate calculation model.
[0051] refer to Figure 3 , Figure 3The flowchart of generating the model configuration table of the items to be processed according to the exemplary embodiment of the present disclosure is schematically shown. In step S310, before generating the model configuration table of the items to be processed, the configuration granularity of the items to be processed can be obtained first. In step S320, after obtaining the configuration granularity of the items to be processed, at least one candidate calculation model corresponding to the items to be processed can be determined according to the configuration granularity. For example, when the configuration type of the items to be processed is configured according to the item granularity, at least one candidate calculation model corresponding to the items to be processed can be determined according to the item information of the items to be processed; when the configuration type of the items to be processed is configured according to the type granularity of the item type corresponding to the items to be processed, the item type corresponding to the items to be processed can be obtained, and at least one candidate calculation model corresponding to multiple items under this item type is the same. For example, if the item to be processed is "tissue paper", and the item category corresponding to "tissue paper" is "daily necessities", all items belonging to "daily necessities" can be processed using the same at least one candidate calculation model. In step S330, after determining at least one candidate computing model corresponding to the object to be processed, a model configuration table of the object to be processed may be generated according to the configuration type of the object to be processed, the object information, and the model information of the candidate computing model.
[0052] It should be noted that when determining the candidate algorithm model corresponding to the item to be processed, the conclusion data, business requirements, etc. of the item to be processed can be processed according to the candidate algorithm model, and the present disclosure does not make any special limitations on this.
[0053] Through the configuration process of the above-mentioned model configuration table, a model configuration table for the item to be processed can be generated. For example, referring to Table 1, Table 1 exemplifies the various fields in the model configuration table. For item 1, the item number corresponding to item 1 is "item 1", and the item number of the item to be processed is unique. The configuration type corresponding to item 1 can be to configure the model configuration table according to the item granularity, and at least one candidate calculation model corresponding to item 1 can be determined according to the item information of item 1, including algorithm A, algorithm B, and algorithm C. The verification modes corresponding to algorithms A, B, and C can be automatic, automatic, and automatic, that is, when processing item 1, it can be dynamically determined whether algorithms A, B, and C match item 1 in the time period to be predicted. The fusion priorities of algorithms A, B, and C can be 1, 10, and 10, respectively, then the fusion priorities of algorithms B and C are the same, and are higher than the fusion priority of algorithm A, and the fusion categories of algorithms A, B, and C can all be mean fusion. For item 2, the item number corresponding to item 2 is "item 2", and the configuration type corresponding to item 2 can be a configuration table based on the item granularity configuration model. According to the item information of item 2, the candidate calculation model corresponding to item 2 can be determined, which is algorithm C. Since the verification mode of algorithm C for item 2 is fixed, when processing item 2, algorithm C will not be dynamically verified, and algorithm C is specified to process item 2. The fusion priority of algorithm C can be 1, and the fusion category can be proportional fusion. In addition, for items belonging to "first-level classification 1", the configuration type of item number "first-level classification 1" is configured according to "first-level category granularity", and it can be determined that the candidate algorithm model corresponding to the items belonging to "first-level classification 1" is algorithm A, and the verification mode of algorithm A for items of "first-level classification 1" is automatic, the fusion priority can be 1, and the fusion category can be proportional fusion.
[0054] Table 1
[0055]
[0056] It should be noted that the above process of configuring the model configuration table can also be performed by directly inserting the model configuration table, obtaining it from the user input interface provided by the system, and synchronizing it with other data sources. The present disclosure does not impose any special limitation on the configuration method of the model configuration table.
[0057] In step S120, a time period to be predicted for the object to be processed is determined, and at least one target computing model matching the object to be processed is determined from at least one candidate computing model according to the time period to be predicted.
[0058] In some exemplary embodiments of the present disclosure, the period to be predicted may be a time interval for predicting the items to be processed, for example, the period to be predicted may be from March 1, 2020 to March 5, 2020. The target computing model may be a computing model that can predict the items to be processed within the period to be predicted.
[0059] After determining the items to be processed, the time period to be predicted for the items to be processed can be further determined, and at least one target calculation model can be determined from at least one candidate calculation model based on the time period to be predicted and the item information of the items to be processed, so as to perform prediction processing on the items to be processed through the target calculation model.
[0060] It is easy for those skilled in the art to understand that, according to specific business needs, the target computing model corresponding to the item to be processed can also be determined in other ways. For example, "item + warehouse" can be used to determine, for example, the target computing model corresponding to item B in warehouse A. The present disclosure does not impose any special limitation on the matching conditions for determining the target computing model of the item to be processed.
[0061] According to some exemplary embodiments of the present disclosure, the following operation steps are performed one by one on the candidate computing models: obtaining and determining the verification mode of the candidate computing model; if the verification mode is a fixed verification mode, obtaining a preconfigured computing model corresponding to the item to be processed, and using the preconfigured computing model as the target computing model; if the verification mode is an automatic verification mode, determining whether the candidate computing model matches the item to be processed based on the time period to be predicted and the item information of the item to be processed; if the candidate computing model matches the item to be processed, determining the candidate computing model as the target computing model.
[0062] refer to Figure 4 , Figure 4 The flowchart for determining the target computing model corresponding to the item to be processed according to the exemplary embodiment of the present disclosure is schematically shown. In step S410, after determining the candidate computing model corresponding to the item to be processed from the model configuration table of the item to be processed, the verification mode of the candidate computing model can be obtained, and the specific mode of the verification mode can be determined. In step S420, if the verification mode is a fixed dynamic verification mode, the preconfigured computing model corresponding to the item to be processed is obtained, and the computing model is used as the target computing model. For example, referring to Table 1, the verification mode corresponding to item 2 is a fixed mode, and the preconfigured computing model corresponding to item 2 is algorithm C. Therefore, algorithm C can be used as the target computing model for item 2. In step S430, if the verification mode is an automatic verification mode, that is, it is possible to dynamically verify whether the candidate computing model matches the item to be processed, then it is determined whether the candidate computing model matches the item to be processed based on the time period to be predicted and the item information of the item to be processed. Reference Figure 2, the relevant historical data 221 of the items to be processed and the basic data 222 of the items to be processed, such as the predicted time period and item information, are input into the dynamic verification module 220, and the dynamic verification module 220 determines whether the candidate calculation model matches the items to be processed in the predicted time period; specifically, the dynamic verification module 220 can verify the candidate calculation model according to the historical sales quantity, historical selling price, promotion rules, external influencing factors and other information of the items to be processed. For example, for item 3, item 3 can be predicted and calculated using the big promotion algorithm model, but there is no big promotion sales strategy in the predicted time period of item 3. Therefore, the dynamic verification module 220 will filter out the big promotion algorithm model, and the big promotion algorithm model will not be the target calculation model corresponding to the items to be processed in the predicted time period. In step S440, if the candidate calculation model matches the items to be processed in the predicted time period, the candidate calculation model is determined as the target calculation model.
[0063] It should be noted that a baseline algorithm model is configured in the present disclosure. When the items to be processed do not match any other calculation models except the baseline algorithm model within the prediction period, the baseline algorithm model will be used to perform prediction processing on the items to be processed, that is, the baseline algorithm model can be used as a fallback target calculation model.
[0064] In step S130, if there are multiple target calculation models, the execution data of the object to be processed is input into each target calculation model respectively, and each target calculation model outputs the calculation result of the object to be processed.
[0065] In some exemplary embodiments of the present disclosure, the calculation result may be a calculation result output by the target calculation model after data such as item information of the item to be processed is input into the target calculation model. Figure 2 If it is determined that there are multiple target calculation models, the distribution module 230 inputs the execution data 241 such as the item information of the item to be processed and the time period to be predicted into each target calculation model, and each target calculation model in the prediction module 240 performs prediction calculation on the item information of the item to be processed, and outputs the corresponding calculation result. For example, according to the algorithm model name of the calculation model matched by the item to be processed, and the mapping relationship between the algorithm model name and the algorithm model library, the execution data of the item to be processed can be input into the corresponding target algorithm model, and each target algorithm model processes the item to be processed according to the execution data.
[0066] After multiple target calculation models corresponding to the items to be processed are determined according to the time period to be predicted, the prediction results of the items to be processed in the time period to be predicted can be obtained according to the calculation results of the multiple target calculation models, for example, the predicted sales volume of the items to be processed in a certain time period in the future. Since different calculation models have their own model characteristics, the lengths of the prediction results obtained by different target calculation models may be inconsistent. For example, when predicting the daily sales volume of item A to be processed between May 20, 2020 and July 20, 2020, the algorithm model matching the item A to be processed may include a big promotion algorithm model, a short-term calculation model, etc.; wherein the big promotion algorithm model is suitable for sales forecast calculation in special scenarios; and the short-term calculation model can be used to calculate the sales volume of a certain item in the next 7 days. For example, the calculation results that can be obtained by the big promotion algorithm model only include the values between June 1, 2020 and June 18, 2020, and the calculation results obtained by using a short-term calculation model may include the values between May 20, 2020 and May 26, 2020. In addition, according to calculation requirements, the short-term calculation model can be used multiple times to calculate the items to be processed. For example, the short-term calculation model outputs the daily sales volume of the item A to be processed between May 20, 2020 and July 20, 2020.
[0067] In step S140, the fusion priority and fusion category of each target calculation model are determined respectively, so as to perform fusion processing on multiple calculation results according to each fusion priority and each fusion category to obtain the prediction result of the object to be processed.
[0068] In some exemplary embodiments of the present disclosure, the fusion process may be a process of performing weighted calculation or mean calculation on multiple calculation results. The prediction result may be a calculation result obtained after the target calculation model is used to process the item to be processed. When the number of target calculation models is determined to be multiple according to the time period to be predicted, after obtaining the calculation results of each target calculation model, the fusion priority and fusion type of each target calculation model may be determined respectively, and the obtained multiple calculation results may be fused according to the fusion priority and fusion type to obtain the prediction result of the item to be processed. Figure 2 After the prediction module 240 outputs the calculation results of multiple target calculation models, the multiple calculation results can be input into the prediction result fusion module 250, and the prediction result fusion module 250 fuses the multiple calculation results to obtain the prediction results of the items to be processed.
[0069] It is easy for a person skilled in the art to understand that when the number of target calculation models is one, the calculation result output by the target calculation model is the prediction result of the item to be processed.
[0070] According to some exemplary embodiments of the present disclosure, the time period to be predicted is divided into multiple unit calculation days, and the unit calculation results corresponding to each unit calculation day are determined; each unit calculation result is fused according to each fusion priority and each fusion category to obtain a unit prediction result; all unit prediction results in the time period to be predicted are used as prediction results. The unit calculation day can be the minimum calculation period divided according to the calculation requirements of the items to be processed. For example, the unit calculation day can be set to one day, two days, etc. The unit calculation result can be the calculation result of the item to be processed calculated by the target calculation model according to the unit calculation day. The unit prediction result can be the result obtained after fusion processing based on the unit calculation result.
[0071] Before fusing multiple calculation results, the time period to be predicted can be divided into multiple unit calculation days, and the unit calculation results corresponding to each target calculation model on each unit calculation day can be obtained respectively. For example, the unit calculation day can be determined as one day, and then the calculation results of each target calculation model can be arranged in date order, and the calculation results corresponding to each target calculation model on each day can be obtained. After obtaining the unit calculation results of each target calculation model, the unit calculation results can be fused according to the fusion priority and fusion category of each target calculation model to obtain the prediction results of the items to be processed. For example, the prediction values of different target calculation models on the same day can be fused to obtain the prediction results corresponding to the items to be processed on this day.
[0072] Table 2
[0073]
[0074] Referring to Table 2, Table 2 schematically shows the relevant calculation data when the data processing method based on multi-model calculation fusion of the present disclosure is used to process items A and B. Table 2 shows that the period to be predicted can be from March 1, 2020 to March 5, 2020. During the period to be predicted, the target calculation model matching item A includes algorithm A, algorithm B, and algorithm C, and the target calculation model matching item B includes algorithm A and algorithm C.
[0075] According to some exemplary embodiments of the present disclosure, each fusion priority is sorted to obtain a level sorting result; the highest priority is determined from the level sorting result; wherein the highest priority is the fusion priority with the largest priority value; if the number of the highest priority is one, the unit calculation result of the target calculation model with the highest priority is used as the unit prediction result. The level sorting result may be a priority sorting result obtained by sorting the fusion priorities of different target calculation models. The highest priority may be the fusion priority with the largest priority value among multiple fusion priorities.
[0076] refer to Figure 5 , Figure 5 A flowchart for determining the prediction results of an item to be processed according to some exemplary embodiments of the present disclosure is schematically shown. In step S510, the fusion priorities of multiple target calculation models are sorted to obtain a level sorting result of the fusion priorities. For example, referring to Table 2, for item B, the fusion priorities of the target calculation model algorithms A and C corresponding to item B can be sorted to obtain the level sorting results of algorithm C and algorithm A. In step S520, the highest priority can be determined from the level sorting results. The highest priority includes 1, which is the fusion priority corresponding to algorithm C (i.e., 10). In step S530, since the number of the highest priority is one, the unit calculation result of the target calculation model corresponding to the highest priority can be used as the unit prediction result.
[0077] According to another exemplary embodiment of the present disclosure, if there are multiple highest priorities, the unit calculation results of the target calculation models corresponding to the multiple highest priorities are used as the results to be fused; and the results to be fused are fused according to the fusion category to obtain the unit prediction result. The result to be fused can be the calculation results output by the target calculation models corresponding to the multiple highest priorities when there are multiple highest priorities.
[0078] refer to Figure 6 , Figure 6 A flowchart for determining the prediction results of an item to be processed according to another exemplary embodiment of the present disclosure is schematically shown. In step S610, when the number of highest priorities is multiple, the unit calculation results of the target calculation models corresponding to the multiple highest priorities can be determined respectively, and the multiple unit calculation results can be used as the results to be fused. Referring to Table 2, for item A, the fusion priorities of the target calculation model algorithms A, B, and C corresponding to item A can be sorted, and the level sorting results obtained are algorithm B, algorithm C, and algorithm A. In step S620, the highest priority can be determined from the level sorting results, and the highest priority includes 2, which are the fusion priorities corresponding to algorithm B and algorithm C. The fusion category of the target algorithm model is obtained, and the unit calculation results are fused according to the obtained fusion category to obtain the unit prediction result.
[0079] According to some exemplary embodiments of the present disclosure, if the fusion category is mean fusion, then a plurality of results to be fused are averaged to obtain a unit prediction result; if the fusion category is proportional fusion, then the calculation weights of each result to be fused are determined, and the unit prediction result is obtained according to each calculation weight and each result to be fused; wherein; determining the calculation weights of each result to be fused includes: determining the calculation error of each target calculation model, and determining the calculation weight of each result to be fused according to each calculation error. Mean fusion may be a fusion processing method for averaging a plurality of unit calculation results. Mean processing may be a processing process for calculating the average of a plurality of unit calculation results. Proportional fusion may be a fusion processing method for performing weighted calculation according to the calculation weights of a plurality of unit calculation results. The calculation weight may be the weight corresponding to each result to be fused when predicting the result. The calculation error may be the error generated by each target calculation model when performing data processing based on multi-model calculation fusion.
[0080] refer to Figure 7 , Figure 7 The flowchart of determining the prediction result of the item to be processed according to the fusion category according to the exemplary embodiment of the present disclosure is schematically shown. In step S710, if the fusion category of the target calculation model is mean fusion, the result to be fused output by the target calculation model corresponding to the highest priority is obtained, and the multiple results to be fused are averaged to obtain the unit prediction result of the item to be processed. For example, the unit prediction result can be Among them, x i is the unit calculation result value of each target calculation model, and N is the number of target calculation models. In step S720, if the fusion category of the target calculation model is proportional fusion, multiple results to be fused are obtained, and the calculation weights of the multiple results to be fused when performing proportional fusion processing are determined. The multiple results to be fused are weighted according to the calculation weights to obtain a unit prediction result. Specifically, the unit prediction result can be determined according to the calculation error of each target calculation model. For example, the unit prediction result can be Among them, r i The calculation error rate of the model for each target calculation, x i is the unit calculation result value of each target calculation model, and N is the number of target calculation models.
[0081] For example, as shown in Table 3, Table 3 schematically shows the prediction results of the items to be processed obtained according to the data processing method based on multi-model computing fusion disclosed in the present invention. Taking the prediction result of item A on March 1, 2020 as an example, since the fusion category of algorithm B and algorithm C is mean fusion, the prediction result of item A on March 1, 2020 is (algorithm B prediction value + algorithm C prediction value) / 2. The prediction result of item A may include all unit prediction results between March 1, 2020 and March 5, 2020.
[0082] Table 3
[0083]
[0084] In summary, the data processing method based on multi-model computational fusion disclosed in the present invention obtains a model configuration table of items to be processed; wherein the model configuration table includes at least one candidate computational model corresponding to the items to be processed; determines the time period to be predicted for the items to be processed, and determines at least one target computational model that matches the items to be processed from at least one candidate computational model according to the time period to be predicted; if there are multiple target computational models, the execution data of the items to be processed are respectively input into each target computational model, and each target computational model outputs the computational result of the items to be processed; respectively determines the fusion priority and fusion category of each target computational model, so as to fuse the multiple computational results according to each fusion priority and each fusion category, and obtains the prediction result of the items to be processed. Through the data processing method based on multi-model computational fusion disclosed in the present invention, on the one hand, the target computational model that matches the time period to be predicted is determined from the candidate computational models, and the objects to be processed are computationally processed by the determined target computational model, which can greatly reduce the resources and time consumed by the execution of each algorithm. On the other hand, when using multiple target calculation models to process the items to be processed, the corresponding calculation model can be dynamically adjusted according to finer-grained constraints such as the time period to be predicted to improve the accuracy of the calculation results. On the other hand, decoupling the personalized adaptation calculation model and fusion category of the item from the standard prediction process can be easily expanded and fine-tuned. On the other hand, when there is a new algorithm model for step-by-step cutting, the range of cutting can be flexibly controlled through the configuration module.
[0085] It should be noted that, although the steps of the method of the present invention are described in a specific order in the drawings, this does not require or imply that the steps must be performed in this specific order, or that all the steps shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps, etc.
[0086] In addition, in this exemplary embodiment, a data processing device based on multi-model computing fusion is also provided. Figure 8 The data processing device 800 based on multi-model computational fusion may include: an information acquisition module 810 , a model determination module 820 , a computation module 830 and a result determination module 840 .
[0087] Specifically, the information acquisition module 810 can be used to obtain a model configuration table for the items to be processed; wherein the model configuration table includes at least one candidate computing model corresponding to the items to be processed; the model determination module 820 can be used to determine the time period to be predicted for the items to be processed, and determine at least one target computing model that matches the items to be processed from at least one candidate computing model based on the time period to be predicted; the computing module 830 can be used to input the execution data of the items to be processed into each target computing model respectively if there are multiple target computing models, and each target computing model outputs the computing result of the items to be processed; the result determination module 840 can be used to respectively determine the fusion priority and fusion category of each target computing model, so as to fuse multiple computing results according to each fusion priority and each fusion category to obtain a prediction result of the items to be processed.
[0088] The data processing device 800 based on multi-model computational fusion can determine the target computational model that matches the time period to be predicted from the candidate computational models, and can greatly reduce the resources and time consumed by the execution of each algorithm by performing computational processing on the items to be processed through the determined target computational model; dynamically adjust the fusion category at a finer granularity according to the item information and the time period to be predicted, which can increase the flexibility of the method for determining the calculation result; decouple the personalized adaptation computational model and fusion category of the items to be processed from the standard prediction process, which can be easily expanded and finely controlled. It is an effective data processing device based on multi-model computational fusion.
[0089] In an exemplary embodiment of the present disclosure, the data processing device based on multi-model computing fusion also includes a model configuration table generation module, which is used to obtain item information of the item to be processed; determine at least one candidate computing model based on the item information, and determine the model information of at least one candidate computing model; generate a model configuration table of the item to be processed based on the determined model information and item information.
[0090] In an exemplary embodiment of the present disclosure, the model determination module includes a model determination unit, which is used to obtain and determine the verification mode of the candidate computing model; if the verification mode is a fixed verification mode, a preconfigured computing model corresponding to the object to be processed is obtained, and the preconfigured computing model is used as the target computing model; if the verification mode is an automatic verification mode, whether the candidate computing model matches the object to be processed is determined based on the prediction period and the object information of the object to be processed; if the candidate computing model matches the object to be processed, the candidate computing model is determined as the target computing model.
[0091] In an exemplary embodiment of the present disclosure, the result determination module includes a result determination unit, which is used to divide the time period to be predicted into multiple unit calculation days, determine the unit calculation results corresponding to each unit calculation day; fuse each unit calculation result according to each fusion priority and each fusion category to obtain a unit prediction result; and use all the unit prediction results in the time period to be predicted as the prediction result.
[0092] In an exemplary embodiment of the present disclosure, the result determination module includes a first result determination unit, which is used to sort each fusion priority to obtain a level sorting result; determine the highest priority from the level sorting result; wherein the highest priority is the fusion priority with the largest priority value; if the number of highest priorities is one, then the unit calculation result of the target calculation model with the highest priority is used as the unit prediction result.
[0093] In an exemplary embodiment of the present disclosure, the result determination unit also includes a second result determination sub-unit, which is used to, if there is a plurality of highest priority levels, use the unit calculation results of the target calculation models corresponding to the plurality of highest priority levels as the results to be fused; and perform fusion processing on each of the results to be fused according to the fusion category to obtain a unit prediction result.
[0094] In an exemplary embodiment of the present disclosure, the result determination unit includes a result determination subunit, which is used to perform mean processing on multiple results to be fused to obtain a unit prediction result if the fusion category is mean fusion; if the fusion category is proportional fusion, determine the calculation weights of each result to be fused, and obtain the unit prediction result according to each calculation weight and each result to be fused; wherein; determining the calculation weight of each result to be fused includes: determining the calculation error of each target calculation model, and determining the calculation weight of each result to be fused according to each calculation error.
[0095] The specific details of the virtual modules of each of the above-mentioned data processing devices based on multi-model computing fusion have been described in detail in the corresponding data processing methods based on multi-model computing fusion, so they will not be repeated here.
[0096] It should be noted that although several modules or units of the data processing device based on multi-model computing fusion are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiment of the present disclosure, the features and functions of two or more modules or units described above can be concretized in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided into multiple modules or units to be concretized.
[0097] In addition, in an exemplary embodiment of the present disclosure, an electronic device capable of implementing the above method is also provided.
[0098] It will be appreciated by those skilled in the art that various aspects of the present invention may be implemented as systems, methods or program products. Therefore, various aspects of the present invention may be specifically implemented in the following forms, namely: complete hardware embodiments, complete software embodiments (including firmware, microcode, etc.), or embodiments combining hardware and software aspects, which may be collectively referred to herein as "circuits", "modules" or "systems".
[0099] Reference below Fig. 9 An electronic device 900 according to such an embodiment of the present invention will be described. Fig. 9 The electronic device 900 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.
[0100] like Fig. 9 As shown, the electronic device 900 is in the form of a general computing device. The components of the electronic device 900 may include, but are not limited to: the at least one processing unit 910, the at least one storage unit 920, a bus 930 connecting different system components (including the storage unit 920 and the processing unit 910), and a display unit 940.
[0101] The storage unit stores program codes, which can be executed by the processing unit 910, so that the processing unit 910 executes the steps according to various exemplary embodiments of the present invention described in the above “Exemplary Method” section of this specification.
[0102] The storage unit 920 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 921 and / or a cache memory unit 922 , and may further include a read-only memory unit (ROM) 923 .
[0103] The storage unit 920 may include a program / utility 924 having a set (at least one) of program modules 925, such program modules 925 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.
[0104] Bus 930 may represent one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.
[0105] The electronic device 900 may also communicate with one or more external devices 970 (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 900, and / or communicate with any device that enables the electronic device 900 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed via an input / output (I / O) interface 950. Furthermore, the electronic device 900 may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via a network adapter 960. As shown, the network adapter 960 communicates with other modules of the electronic device 900 via a bus 930. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 900, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.
[0106] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the embodiment of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the embodiment of the present disclosure.
[0107] In an exemplary embodiment of the present disclosure, a computer-readable storage medium is also provided, on which a program product capable of implementing the above method of the present specification is stored. In some possible embodiments, various aspects of the present invention can also be implemented in the form of a program product, which includes a program code, and when the program product is run on a terminal device, the program code is used to enable the terminal device to perform the steps according to various exemplary embodiments of the present invention described in the above "Exemplary Method" section of the present specification.
[0108] refer to Fig.10 As shown, a program product 1000 for implementing the above method according to an embodiment of the present invention is described, which can adopt a portable compact disk read-only memory (CD-ROM) and include program code, and can be run on a terminal device, such as a personal computer. However, the program product of the present invention is not limited thereto. In this document, a readable storage medium can be any tangible medium containing or storing a program, which can be used by or in combination with an instruction execution system, an apparatus or a device.
[0109] The program product may use any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.
[0110] Computer readable signal media may include data signals propagated in baseband or as part of a carrier wave, in which readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Readable signal media may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.
[0111] The program code embodied on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the foregoing.
[0112] Program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages such as Java, C++, etc., and conventional procedural programming languages such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).
[0113] In addition, the above-mentioned figures are only schematic illustrations of the processes included in the method according to an exemplary embodiment of the present invention, and are not intended to be limiting. It is easy to understand that the processes shown in the above-mentioned figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.
[0114] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary technical means in the art that are not disclosed in the present disclosure. The specification and examples are to be considered as exemplary only, and the true scope and spirit of the present disclosure are indicated by the claims.
[0115] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.
Claims
1. A data processing method based on multi-model computing fusion, It is characterized in that include: Obtaining a model configuration table of an object to be processed; wherein the model configuration table includes at least one candidate calculation model corresponding to the object to be processed; Determine a time period to be predicted for the object to be processed, and determine at least one target computing model matching the object to be processed from the at least one candidate computing model according to the time period to be predicted; There are multiple target calculation models, and the execution data of the object to be processed is input into each of the target calculation models respectively, and each of the target calculation models outputs the calculation result of the object to be processed; Determining the fusion priority and fusion category of each of the target calculation models respectively, so as to perform fusion processing on the plurality of calculation results according to each of the fusion priorities and each of the fusion categories, and obtain the prediction result of the object to be processed; The step of determining at least one target computing model matching the object to be processed from the at least one candidate computing model according to the time period to be predicted includes: Perform the following steps on each candidate computing model: Obtaining and determining a verification mode of the candidate computing model; If the verification mode is a fixed verification mode, obtaining a preconfigured calculation model corresponding to the object to be processed, and using the preconfigured calculation model as the target calculation model; If the verification mode is an automatic verification mode, judging whether the candidate calculation model matches the object to be processed according to the time period to be predicted and the object information of the object to be processed; If the candidate computing model matches the object to be processed, determining the candidate computing model as the target computing model; The fusing the plurality of calculation results according to the fusion priorities and the fusion categories to obtain the prediction results of the items to be processed includes: Dividing the time period to be predicted into a plurality of unit calculation days, and determining the unit calculation results corresponding to each of the unit calculation days; Performing fusion processing on each of the unit calculation results according to each of the fusion priorities and each of the fusion categories to obtain a unit prediction result; Taking all unit prediction results in the time period to be predicted as the prediction result; If the fusion category is mean fusion, then a plurality of results to be fused are processed by averaging to obtain the unit prediction result; If the fusion category is proportional fusion, the calculation weights of the results to be fused are determined, and the unit prediction result is obtained according to the calculation weights and the results to be fused.
2. The data processing method based on multi-model computing fusion according to claim 1, It is characterized in that Before obtaining the model configuration table of the items to be processed, the method further includes: Acquire item information of the item to be processed; Determine the at least one candidate computing model according to the item information, and determine model information of the at least one candidate computing model; A model configuration table of the object to be processed is generated according to the determined model information and the object information.
3. The data processing method based on multi-model computing fusion according to claim 1, It is characterized in that The fusing the unit calculation results according to the fusion priorities and the fusion categories to obtain the unit prediction results includes: Sorting the fusion priorities to obtain a level sorting result; Determine the highest priority from the level sorting results; wherein the highest priority is the fusion priority with the largest priority value; If the number of the highest priority is one, the unit calculation result of the target calculation model with the highest priority is used as the unit prediction result.
4. The data processing method based on multi-model computing fusion according to claim 3, It is characterized in that The fusing the unit calculation results according to the fusion priorities and the fusion categories to obtain unit prediction results also includes: If there are multiple highest priorities, the unit calculation results of the target calculation models respectively corresponding to the multiple highest priorities are used as the results to be fused; The results to be fused are fused according to the fusion category to obtain the unit prediction result.
5. The data processing method based on multi-model computing fusion according to claim 1, It is characterized in that Determining the calculation weights of the results to be fused includes: The calculation errors of the target calculation models are determined, and the calculation weights of the results to be fused are determined according to the calculation errors.
6. A data processing device based on multi-model computing fusion, It is characterized in that include: An information acquisition module, used to acquire a model configuration table of the object to be processed; wherein the model configuration table includes at least one candidate calculation model corresponding to the object to be processed; A model determination module, used to determine a time period to be predicted for the object to be processed, and determine at least one target computing model matching the object to be processed from the at least one candidate computing model according to the time period to be predicted; A calculation module, for a plurality of target calculation models, inputting the execution data of the object to be processed into each of the target calculation models respectively, and outputting the calculation result of the object to be processed by each of the target calculation models; A result determination module, used to respectively determine the fusion priority and fusion category of each of the target calculation models, so as to perform fusion processing on the multiple calculation results according to each of the fusion priorities and each of the fusion categories to obtain the prediction result of the item to be processed; The model determination module is further used to perform the following operation steps on the candidate computing models one by one: obtaining and determining the verification mode of the candidate computing model; If the verification mode is a fixed verification mode, obtaining a preconfigured calculation model corresponding to the object to be processed, and using the preconfigured calculation model as the target calculation model; If the verification mode is an automatic verification mode, judging whether the candidate calculation model matches the object to be processed according to the time period to be predicted and the object information of the object to be processed; If the candidate computing model matches the object to be processed, determining the candidate computing model as the target computing model; The result determination module is further used to divide the time period to be predicted into a plurality of unit calculation days, and determine the unit calculation results corresponding to each of the unit calculation days; Performing fusion processing on each of the unit calculation results according to each of the fusion priorities and each of the fusion categories to obtain a unit prediction result; Taking all unit prediction results in the time period to be predicted as the prediction result; If the fusion category is mean fusion, then a plurality of results to be fused are processed by averaging to obtain the unit prediction result; If the fusion category is proportional fusion, the calculation weights of the results to be fused are determined, and the unit prediction result is obtained according to the calculation weights and the results to be fused.
7. An electronic device, It is characterized in that include: processor; as well as A memory having computer-readable instructions stored thereon, wherein the computer-readable instructions, when executed by the processor, implement the data processing method based on multi-model computational fusion according to any one of claims 1 to 5.
8. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the data processing method based on multi-model computational fusion according to any one of claims 1 to 5 is implemented.
Citation Information
Patent Citations
Maize yield combined prediction system and method
CN104732299A
Goods sales volume prediction method and device, and electronic equipment
CN106408341A
Method for constructing sequence forecasting model based on multi-visual-angle data and cycle network
CN106600347A
Multivariable time series prediction method and system
CN107146015A
Merchant passenger flow big data prediction method based on time series decomposition
CN107895283A
