Attribution analysis method and device, electronic equipment, storage medium and program product

Abstract

The invention provides an attribution analysis method and device, electronic equipment, a storage medium and a program product, and relates to the technical field of data processing. According to the method, related index configuration parameters are configured through a user, analysis is automatically achieved through a system, efficiency and accuracy are higher, attribution analysis of different attribution analysis types and different attribution dimension organization forms can be supported, flexibility is higher, and different analysis requirements of the user can be met.

Description

Technical Field

[0001] The present application relates to the technical field of data processing, and in particular, to an attribution analysis method, device, electronic device, storage medium, and program product. Background Art

[0002] With the popularization of the Internet, companies in all walks of life are producing and accumulating a huge amount of data every moment. In order to utilize this data, different companies have extracted a large number of different data indicators from the data according to the company's business preferences to measure the company's business development, product quality, etc. Among them, abnormal indicators refer to the overall data anomalies caused by certain potential factors.

[0003] The purpose of attribution analysis is to find out the data dimensions or factors that cause or have a greater impact on anomalies. Currently, it relies on manual work to locate the root cause of an abnormal or fluctuating abnormal indicator, which is very time-consuming and difficult to ensure accuracy. Summary of the Invention

[0004] The purpose of the embodiments of the present application is to provide an attribution analysis method, device, electronic device, storage medium, and program product to improve the problems of low efficiency and low accuracy of the existing analysis methods.

[0005] In a first aspect, the embodiments of the present application provide an attribution analysis method, and the method includes:

[0006] Determine the attribution index to be analyzed and related index configuration parameters, where the related index configuration parameters include the attribution analysis type configured by the user, the attribution dimension organization form, and the attribution dimension. The attribution analysis type includes dimension achievement attribution and dimension fluctuation attribution, and the attribution dimension organization form includes single dimension, combined dimension, and dimension drilling down;

[0007] According to the related index configuration parameters, calculate the influence degree of each attribution dimension on the attribution index to be analyzed;

[0008] Determine the attribution analysis result according to the influence degree.

[0009] In the above implementation process, by allowing the user to configure the related index configuration parameters and automatically implementing the analysis by the system, the efficiency and accuracy are higher, and it can support attribution analysis of different attribution analysis types and different attribution dimension organization forms, with higher flexibility and can meet the different analysis needs of users.

[0010] Optionally, the calculating the influence degree of each attribution dimension on the attribution index to be analyzed according to the related index configuration parameters includes:

[0011] Traverse the constructed data tree according to the attribution dimension in the relevant index configuration parameters, where each node in the data tree includes the dimension name, dimension value, index value of the attribution dimension, and child nodes;

[0012] For each node traversed, calculate the influence degree of the corresponding attribution dimension of the node on the attribution index to be analyzed.

[0013] In the above implementation process, by traversing the data tree for dimension search, the influence degree of each attribution dimension on the index can be comprehensively analyzed. By calculating the influence degree of each node, the dimension combination with the greatest influence on the attribution index can be quickly identified.

[0014] Optionally, the data tree is constructed in the following manner:

[0015] Obtain the raw data to be analyzed from the data source according to the attribution index to be analyzed and the attribution dimension in the relevant index configuration parameters, where the raw data includes multiple attribution dimensions and corresponding index values;

[0016] Perform aggregation processing on the raw data according to the preset dimension combination rules to generate aggregated data at different levels, and each level of aggregated data corresponds to a combination of one or more attribution dimensions;

[0017] Generate a hierarchical data tree based on the aggregated data at different levels.

[0018] In the above implementation process, the data tree organizes data through a hierarchical structure, enabling quick positioning to the target node during the search process.

[0019] Optionally, the relevant index configuration parameters further include the data date range configured by the user. Obtaining the raw data to be analyzed from the data source according to the attribution index to be analyzed and the attribution dimension in the relevant index configuration parameters includes:

[0020] Obtain the raw data to be analyzed from the data source according to the attribution index to be analyzed, the attribution dimension in the relevant index configuration parameters, and the data date range. In this way, the data within the required analysis date range can be flexibly selected according to the needs.

[0021] Optionally, traversing the constructed data tree according to the attribution dimension in the relevant index configuration parameters includes:

[0022] Traverse the constructed data tree using the depth-first search algorithm according to the attribution dimension in the relevant index configuration parameters.

[0023] In the above implementation process, all possible dimension combinations can be efficiently traversed through the depth-first search algorithm, and the nodes that have a significant impact on the change of the metric can be quickly found.

[0024] Optionally, after calculating the degree of influence of the attribution dimension corresponding to each traversed node on the attribution metric to be analyzed, the method further includes:

[0025] Pruning the traversed nodes according to a preset pruning strategy, and putting the nodes that meet the preset pruning strategy into a stack as candidate search nodes for the subsequent depth-first search algorithm.

[0026] In the above implementation process, the combined use of the stack and the pruning strategy can efficiently screen out the dimension combinations that have a significant impact on the change of the metric when traversing the data tree, avoiding useless searches.

[0027] Optionally, the preset pruning strategy includes that the degree of influence is greater than a set threshold, or the preset pruning strategy includes that the proportion of the data volume corresponding to the node is greater than a set data volume proportion, or the preset pruning strategy includes a pruning strategy set according to the business logic of attribution analysis. By setting the pruning strategy, the nodes that have little impact on the change of the metric can be pruned during the search process, so that it is possible to avoid continuing to explore the sub-nodes of this node, avoid useless searches, and improve the search efficiency of the algorithm.

[0028] Optionally, the attribution analysis type further includes factor fluctuation attribution. In the case where the attribution analysis type included in the relevant metric configuration parameters is factor fluctuation attribution, the method further includes:

[0029] Determining the type of the metric calculation formula corresponding to the attribution metric to be analyzed;

[0030] Determining the calculation method of the degree of influence according to the type of the metric calculation formula;

[0031] Calculating the degree of influence of each calculation factor in the metric calculation formula on the attribution metric to be analyzed according to the calculation method;

[0032] Determining the attribution analysis result according to the degree of influence.

[0033] In the above implementation process, for factor fluctuation attribution, different calculation methods are selected according to the type of the calculation formula to calculate the degree of influence, which can improve the accuracy of calculating the degree of influence.

[0034] Optionally, after obtaining the degree of influence, the method further includes:

[0035] According to the index attributes of the attribution index to be analyzed, normalize the influence degree to obtain the normalized influence degree, where the index attributes include positive attributes and negative attributes. Normalization can clarify the contribution direction (positive or negative) of each dimension or factor to the index change, so that users can quickly know the root causes with greater influence through the attribution analysis results.

[0036] Second, an embodiment of the present application provides an attribution analysis device, which includes:

[0037] A data acquisition module, configured to determine the attribution index to be analyzed and related index configuration parameters, where the related index configuration parameters include the attribution analysis type configured by the user, the attribution dimension organization form, and the attribution dimension. The attribution analysis type includes dimension achievement attribution and dimension fluctuation attribution. The attribution dimension organization form includes a single dimension, a combined dimension, and dimension drilling down;

[0038] A calculation module, configured to calculate the influence degree of each attribution dimension on the attribution index to be analyzed according to the related index configuration parameters;

[0039] A result determination module, configured to determine the attribution analysis result according to the influence degree.

[0040] Third, an embodiment of the present application provides an electronic device, including a processor and a memory. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the steps in the method provided in the first aspect above are run.

[0041] Fourth, an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps in the method provided in the first aspect above are run.

[0042] Fifth, an embodiment of the present application provides a computer program product, including computer program instructions. When the computer program instructions are read and run by a processor, the steps in the method provided in the first aspect above are executed.

[0043] Other features and advantages of the present application will be described in the subsequent specification, and, in part, will be obvious from the specification, or will be understood by implementing the embodiments of the present application. The objectives and other advantages of the present application can be realized and obtained by the structures specifically pointed out in the written specification, claims, and drawings. Description of the Drawings

[0044] To more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the accompanying drawings required for the embodiments of the present application. It should be understood that the following drawings only show some embodiments of the present application, and therefore should not be regarded as a limitation of the scope. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0045] Figure 1 It is a flowchart of an attribution analysis method provided by an embodiment of the present application;

[0046] Figure 2 It is a schematic diagram of the combination of various attribution dimensions provided by an embodiment of the present application;

[0047] Figure 3 It is a schematic diagram of the structure of a data tree provided by an embodiment of the present application;

[0048] Figure 4 It is a schematic diagram of the tree structure corresponding to the search process of a depth-first search algorithm provided by an embodiment of the present application;

[0049] Figure 5 It is a structural block diagram of an attribution analysis device provided by an embodiment of the present application;

[0050] Figure 6 It is a schematic diagram of the structure of an electronic device for executing the attribution analysis method provided by an embodiment of the present application. Detailed implementation manners

[0051] The following will clearly and completely describe the technical solutions in the embodiments of the present application in conjunction with the accompanying drawings in the embodiments of the present application.

[0052] It should be noted that the terms "system" and "network" in the embodiments of the present invention can be used interchangeably. "Multiple" means two or more. In view of this, in the embodiments of the present invention, "multiple" can also be understood as "at least two". "And / or" describes the association relationship of associated objects and indicates that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " generally represents an "or" relationship between the associated objects before and after without special instructions.

[0053] It should also be noted that all actions of obtaining signals, information, or data in the present application are carried out on the premise of complying with the corresponding data protection regulations and policies of the country where it is located and obtaining the authorization given by the owner of the corresponding device.

[0054] An embodiment of the present application provides an attribution analysis method. This method determines an attribution index to be analyzed and relevant index configuration parameters. The relevant index configuration parameters can be configured by the user and include the attribution analysis type, the attribution dimension organization form, and the attribution dimension. The attribution analysis type includes dimension achievement attribution and dimension fluctuation attribution. The attribution dimension organization form includes single dimension, combined dimension, and dimension drill-down. Then, the influence degree of each attribution dimension on the attribution index to be analyzed can be calculated according to the relevant index configuration parameters, and the attribution analysis result can be determined based on the influence degree. Therefore, in this solution, the user can configure the relevant index configuration parameters, and the system can automatically perform the analysis, with higher efficiency and accuracy. Moreover, it supports attribution analysis of different attribution analysis types and different attribution dimension organization forms, with higher flexibility and better adaptation to the user's needs.

[0055] Please refer to Figure 1 , Figure 1 FIG. is a flowchart of an attribution analysis method provided by an embodiment of the present application. The method includes the following steps:

[0056] Step S110: Determine the attribution index to be analyzed and the relevant index configuration parameters.

[0057] Among them, the attribution analysis method of the present application can be executed by an intelligent analysis system. When analysis is required, the user can configure the attribution index to be analyzed and the relevant index configuration parameters in the intelligent analysis system.

[0058] The attribution index can refer to the business index that needs to be analyzed in the attribution analysis and is a key index for measuring the performance of a certain business activity or process. The attribution analysis can analyze what factors affect this index.

[0059] The attribution index to be analyzed can be selected by the user in the intelligent analysis system according to the analysis requirements. Understandably, the intelligent analysis system can provide various attribution indexes, such as order timeout rate, sales growth rate, customer churn rate, average order value, order conversion rate, product page view volume, product collection volume, etc. For example, if an online sales platform finds that its order timeout rate has increased from 5% to 10% and needs to analyze what factors have caused this change, the user can select the order timeout rate as the attribution index to be analyzed.

[0060] The intelligent analysis system can also provide various attribution analysis types, attribution dimension organization forms, and attribution dimensions for the user to choose. The attribution analysis type can include dimension achievement attribution and dimension fluctuation attribution.

[0061] Dimension achievement attribution disassembles the achievement of an indicator across various dimensions. Through the specific performance of different dimensions, users can understand the degree of influence of each dimension on the achievement of the indicator. This type of attribution analysis is applicable when the indicator value reaches xx, to analyze the influence of each dimension and dimension value on the achievement of this indicator. For example, in scenarios where the order overtime rate is higher than 50%, the prediction accuracy rate is lower than 80%, the daily pv / uv is lower than 1000, etc., this type of attribution analysis can be used for analysis. It can locate which dimensions are the root causes affecting the achievement of the indicator, and then users can adjust according to the corresponding measures for these dimensions to restore the indicator value to the expected level.

[0062] Dimension fluctuation attribution disassembles the change of an indicator across various dimensions and gives a quantitative contribution explanation. Through dimension fluctuation attribution, users can understand the reasons for the change of the overall market indicator and clarify the degree of influence of each dimension on the change of the indicator. The difference from the above-mentioned dimension achievement attribution is that dimension achievement attribution only needs to look at a single period, while dimension fluctuation attribution refers to the abnormal fluctuation of the indicator value in two periods, and it is necessary to analyze which dimensions cause this abnormal fluctuation. It is applicable to scenarios where it is necessary to compare the indicator value fluctuations in two periods. For example, if the order overtime rate fluctuates greatly by more than 20%, it can be analyzed whether it is due to the influence of certain dimensions (such as whether the express pickup is timely, whether there is abnormal weather, etc.).

[0063] In some implementation manners, the attribution analysis type may further include factor fluctuation attribution, which can be used to explore the degree of influence of multiple related factors on the change of an indicator. Factor fluctuation attribution disassembles the abnormal fluctuation to each factor in the indicator calculation formula. For example, the calculation formula of a certain indicator is (A - B) / C, which contains three factors A, B, and C. In two periods, it can be analyzed which factor's change brings a large fluctuation to the indicator.

[0064] The attribution dimension organization form may include a single dimension, dimension combination, and dimension drill-down.

[0065] A single dimension means that only one dimension is looked at during attribution analysis. For example, only looking at the city dimension, the result will be reflected in the specific city. Dimension combination means that during attribution analysis, the selected dimensions will be cross-multiplied to explore whether each dimension combination is the root cause. Dimension drill-down refers to the order of selecting dimensions during attribution analysis, which is the path for the algorithm to explore the root cause. For example, if the selected dimensions include business type, shipping warehouse, and carrier, then it can be explored in the order of business type -> shipping warehouse -> carrier. Generally, the drill-down depth can be configured for dimension drill-down, that is, the maximum number of dimensions included in the output attribution analysis result. Usually, 5 dimensions can reflect most problems, so generally the depth can be set to 5.

[0066] The attribution dimension can also be selected by the user in the intelligent analysis system, that is, the user can determine from which dimensions to analyze the anomalies of the indicators. Usually, they are the dimensions that the business is concerned about. There are some restrictions when choosing different organizational forms of attribution dimensions. For example, in the case of a single dimension, there is no limit to the number of dimension selections; in the case of dimension drilling down, it does not exceed 8; and in the case of dimension combination, it does not exceed 5.

[0067] The attribution dimensions can include city dimension, province dimension, time dimension, business type, express delivery, commodity type, carrier, etc. It can be understood that in different business scenarios, the attribution dimensions can be different.

[0068] Therefore, the user can select information such as the attribution indicators to be analyzed, the attribution analysis type, the organizational form of attribution dimensions, and the attribution dimensions in the intelligent analysis system according to their own analysis needs.

[0069] In some implementation manners, after the user selects the attribution indicators to be analyzed, the intelligent analysis system can recommend relevant attribution dimensions to the user according to the attribution indicators to be analyzed. At this time, the recommended attribution dimensions are some dimensions with a relatively high degree of association with the attribution indicators to be analyzed. In this way, the user can directly select the attribution dimensions recommended by the system for analysis, and the user does not need to judge which attribution dimensions have a higher degree of relevance to the attribution indicators to be analyzed, which is more efficient.

[0070] Step S120: Configure parameters according to relevant indicators and calculate the influence degree of each attribution dimension on the attribution indicator to be analyzed.

[0071] After receiving the attribution indicators to be analyzed selected by the user and the relevant indicator configuration parameters, the intelligent analysis system can calculate the influence degree of each attribution dimension on the attribution indicator to be analyzed according to the relevant indicator configuration parameters.

[0072] In some implementation manners, the influence degree can be characterized by the contribution degree, that is, the contribution degree can be used to quantify the influence degree of different factors or events on the indicator result. Here, when calculating the influence degree, in different attribution analysis types, the way of calculating the contribution degree can be different, and in different organizational forms of attribution dimensions, the number of dimensions calculated can be different. For example, in the case of combined dimensions, it is necessary to calculate the contribution degrees of various combined dimensions. Therefore, the contribution degrees of each attribution dimension can be calculated to be used to evaluate the influence degree on the attribution indicator to be analyzed.

[0073] Step S130: Determine the attribution analysis result according to the influence degree.

[0074] After obtaining the influence degree above, it can be determined according to the influence degree which dimensions have a greater influence on the attribution indicator to be analyzed, that is, the attribution analysis result includes the attribution dimensions with a large influence.

[0075] In some embodiments, the influence degrees of all attribution dimensions can be sorted by size, and then the attribution dimension with the greatest influence on the change of the metric can be identified. For example, when analyzing the order overtime rate, the contribution degree of the business type "fresh food" is 0.4, the contribution degree of "xx warehouse" is 0.3, and the contribution degree of "xx express" is 0.5. When determining the attribution analysis result, these contribution segments can be sorted by size and output and presented to the user, or a threshold can be set, and the dimensions higher than the threshold can be output to the user as the attribution analysis result. For example, if the threshold is 0.3, the contribution degrees of "fresh food" and "xx express" are higher than the threshold, and it can be analyzed that the factors affecting the order overtime rate include "fresh food" and "xx express".

[0076] In the above implementation process, relevant metric configuration parameters are configured by the user, and the analysis is automatically implemented by the system, with higher efficiency and accuracy. Moreover, it supports attribution analysis of different attribution analysis types and different attribution dimension organization forms, with higher flexibility and can meet the different analysis needs of users.

[0077] Based on the above embodiments, in some application scenarios, the dimensions that may affect the metric may be very numerous and there may be many dimension values under a single dimension. Therefore, a large amount of data needs to be searched during the analysis, which is time-consuming. To improve this problem, in this solution, during the process of calculating the influence degree, a data tree structure is used for dimension exploration. Specifically, according to the attribution dimensions in the relevant metric configuration parameters, the constructed data tree can be traversed. Among them, each node in the data includes the dimension name, dimension value, metric value of the attribution dimension, and sub-nodes. Then, for each node traversed, calculate the influence degree of the corresponding attribution dimension of the node on the attribution metric to be analyzed.

[0078] In the data tree, each node has only one dimension of its own, and for the combination of multiple dimensions, look down along the parent node, and one more dimension is added for each layer found until the dimensions correspond. For example, if you want to find the node data with the dimension combination of A, B, and C, the top-level node node has the dimension name = A. Filter out the node with the dimension name = B in the child nodes of node, and then look down for its child nodes and filter out the node with the dimension name = C, and then the dimension combination data that has been aggregated can be directly used for calculation.

[0079] It can be understood that the data tree is a structure presented after statistically integrating the original data. For example, if the attribution metric to be analyzed is the order overtime rate, then the metric values of each node in the constructed data tree include the number of overtime orders, the total number of orders, the overtime rate, etc. In this way, by traversing the data tree, the metric values of the corresponding nodes can be quickly obtained and then the contribution degree can be calculated, improving the calculation efficiency.

[0080] In the above implementation process, traversing through the data tree is used to perform dimensional search, which can comprehensively analyze the influence degree of each attribution dimension on the metric. By calculating the influence degree of each node, the dimension combination that has the greatest impact on the attribution metric can be quickly identified.

[0081] Based on the above embodiments, the data tree can be constructed in the following manner: According to the attribution metric to be analyzed and the attribution dimensions in the relevant metric configuration parameters, the raw data to be analyzed is obtained from the data source. The raw data includes multiple attribution dimensions and corresponding metric values; according to the preset dimension combination rules, the raw data is aggregated to generate aggregated data at different levels, and each level of aggregated data corresponds to a combination of one or more attribution dimensions; then, a hierarchical data tree can be generated according to the aggregated data at different levels.

[0082] According to the attribution metric to be analyzed and the attribution dimensions, the relevant raw data can be retrieved from the data source for constructing the data tree. In a specific implementation manner, the intelligent analysis system can submit the attribution metric to be analyzed and the attribution dimensions to the metric platform. The metric platform can provide more powerful data retrieval capabilities. The metric platform can generate corresponding SQL statements according to the information obtained, and return the SQL statements to the intelligent analysis system. After the intelligent analysis system obtains the SQL statements, it can retrieve from the database to obtain the raw data to be analyzed. It can be understood that the intelligent analysis system can also generate SQL statements by itself for data retrieval.

[0083] When performing data retrieval, the SQL statement will be parsed and transformed, and a groupby cube operation will be performed on the selected attribution dimensions. For example, if the selected attribution dimensions include three dimensions A, B, and C (corresponding to three column names in the data table), the numerator, denominator, and metric value in the calculation formula of the attribution metric to be analyzed (such as the number of overtime orders, the total number of orders, the overtime rate), this operation will generate A, numerator, denominator, metric value, B, numerator, denominator, metric value, C, numerator, denominator, metric value, A, B, numerator, denominator, metric value, A, C, numerator, denominator, metric value, B, C, numerator, denominator, metric value, that is, various combination situations of the selected attribution dimensions, as Figure 2 shown.

[0084] After retrieving the raw data from the database, the dirty data in the original data can be cleaned first to remove invalid values, outliers, and missing values to ensure data quality and avoid interfering with the subsequent algorithm search process.

[0085] Then, according to the preset dimension combination rules, the original data can be aggregated. The preset dimension combination rules here can be any combination of multiple attribution dimensions or combination in a predefined order. The aggregation process may include performing a group by cube operation on the original data to generate all possible dimension combination aggregation results. The metric value contained in each node is the result after aggregation under this dimension combination. The metric value of a certain node may be the average order timeout rate under this dimension combination.

[0086] The aggregation process is equivalent to aggregating and statistically analyzing each attribution dimension and the data under various attribution dimension combinations for presentation in the data tree. When constructing the data tree, the dimension name of a node can represent the dimension at this level, and the dimension value is the specific value of this dimension. Starting from the highest level (the overall market dimension), each level of the tree is gradually constructed. The highest-level node may only have the dimension name "overall market" without a specific dimension value, and its child nodes may be specific dimension values under a certain dimension (such as "business type") (such as "brand direct delivery"). The child nodes of each node contain the dimension combination of the next level. For example, if the parent node is "business type = brand direct delivery", then its child nodes may be "carrier product = express delivery". The constructed data tree structure is as Figure 3 shown Figure 3 where different grayscales are used to represent the elements under different dimensions.

[0087] In the above implementation process, by constructing the data tree, during the algorithm search process, the hierarchical results of the data tree can be utilized to quickly locate the data under the attribution dimension without repeatedly extracting data from the data source, thus greatly shortening the calculation time.

[0088] Based on the above embodiments, the relevant metric configuration parameters may further include the data date range configured by the user. When obtaining the original data, the original data to be analyzed can be obtained from the data source according to the attribution metrics to be analyzed, the attribution dimensions in the relevant metric configuration parameters, and the data date range.

[0089] The data date range can be used to indicate the range of the original data obtained. For example, if the user wants to view the data for a period of time for analysis, the data date range can be indicated, and thus the data to be analyzed can be quickly retrieved according to the requirements.

[0090] It can be understood that if the attribution analysis type is dimension achievement attribution, only one date range needs to be selected. If the attribution analysis type is dimension fluctuation attribution, two-period data ranges, namely the current period and the comparison period, need to be selected.

[0091] Based on the above embodiments, in the case where the attribution type is dimensional fluctuation attribution, since it is necessary to compare the data of two cycles, when constructing the data tree above, a data tree can be constructed for the data of each cycle. When comparing, it is necessary to compare the data of the same dimensional combination in two cycles. Therefore, constructing two data trees can quickly search for the data of different dimensions within one cycle.

[0092] Based on the above embodiments, in the way of traversing the data tree, the constructed data tree can be traversed using the depth-first search algorithm according to the attribution dimension in the configuration parameters of relevant indicators.

[0093] The depth-first search algorithm is an algorithm for traversing or searching a tree or graph. Its implementation principle is based on the mechanisms of recursion and backtracking. The algorithm starts from the root node or any starting node and searches as deeply as possible along the branches of the tree until it can no longer continue. The depth-first search algorithm is usually implemented through a recursive function. The recursive function will visit the current node and recursively call itself for each child node. Each recursive call will process one child node until there are no more child nodes to visit.

[0094] The implementation process of the depth-first search algorithm can also be understood as a tree structure, as Figure 4 shown, Figure 4 where different grayscales are used to represent the elements under different dimensions. When searching, it will explore all the way down a single branch until the drill-down depth is met or there are no more indicators, and then the branch ends.

[0095] In the above implementation process, through the depth-first search algorithm, all possible dimensional combinations can be efficiently traversed, and the nodes that have a significant impact on the indicator change can be quickly found.

[0096] Based on the above embodiments, in order to improve the analysis efficiency, after traversing each node and calculating the influence degree of the node above, the nodes traversed can be pruned according to the preset pruning strategy, and the nodes that meet the preset pruning strategy are put into the stack as the candidate search nodes for the subsequent depth-first search algorithm.

[0097] In the recursive implementation of the depth-first search algorithm, the stack will automatically maintain the access order of the nodes. Each recursive call will push the current node onto the stack and pop the stack after processing.

[0098] Therefore, before pushing the current node onto the stack here, the preset pruning strategy also needs to be applied to the node. If it meets the preset pruning strategy, it will be pushed onto the stack; if not, it will not be pushed onto the stack. In this case, for the nodes that do not meet the requirements, further exploration will not continue.

[0099] In some embodiments, the intelligent analysis system can provide some pruning strategies for users, and users can flexibly choose whether to enable the pruning strategies and which pruning strategy to enable according to their needs. If the pruning strategy is met, it indicates that the dimension has a greater impact on the change of the metric. If the pruning strategy is not met, it indicates that the dimension has a smaller impact on the change of the metric. In this case, there is no need to continue exploring the next dimension, which can avoid processing a large number of irrelevant nodes, reduce the amount of calculation, and improve the search efficiency.

[0100] The data tree is a hierarchical data structure. Each node represents a dimension combination. The root node is usually the overall market dimension, and its child nodes are different dimension values (such as business type, shipping warehouse, etc.). The child nodes of the child nodes are finer-grained dimension combinations (such as business type x shipping warehouse). The depth-first search algorithm is used to traverse the data tree. The search starts from the root node and explores each node along the branches of the tree. When each node is visited, the degree of its impact on the metric change is calculated, and according to the preset pruning strategy, it is decided whether to continue exploring its child nodes.

[0101] The stack is used to store the nodes to be visited. In the implementation process of the depth-first search algorithm, the child nodes of the current node are first pushed into the stack. Generally, for subsequent processing, if the degree of impact of the child node is calculated and the child node does not meet the preset pruning strategy, the child node will be removed from the stack and its child nodes will no longer be processed.

[0102] For example, assume that the data tree is constructed. The root node is the overall market dimension, and its child nodes are business types (brand direct shipping, consignment, fresh food). The child nodes of each business type node are shipping warehouses (Shanghai warehouse, Beijing warehouse). The child nodes of each shipping warehouse node are carriers (XX Express, YY Express).

[0103] Initialize the stack, push the root node into the stack, calculate the degree of impact of the overall market dimension, and push its child nodes (brand direct shipping, consignment, fresh food) into the stack. According to the preset pruning strategy, assume that the impact degree of the "consignment" business type does not meet the preset pruning strategy, and remove it from the stack. Continue to visit the "brand direct shipping" node, calculate the impact degree of "brand direct shipping", and push its child nodes (Shanghai warehouse, Beijing warehouse) into the stack. If the "Shanghai warehouse" does not meet the preset pruning strategy, remove it from the stack according to the preset pruning strategy. Continue to visit the "Beijing warehouse" node, calculate the impact degree of "Beijing warehouse", and push its child nodes (XX Express, YY Express) into the stack. Continue to visit the "XX Express" node, calculate the impact degree of "brand direct shipping × Beijing warehouse × XX Express". If the preset pruning strategy is met, its child nodes (such as the next dimension combination) may be pushed into the stack and continue to be explored in depth.

[0104] In the above implementation process, the combined use of the stack and the pruning strategy can efficiently screen out the dimension combinations that have a significant impact on the metric changes when traversing the data tree.

[0105] Based on the above embodiments, the preset pruning strategy may include that the influence degree is greater than a set threshold, or the preset pruning strategy includes that the proportion of the data volume corresponding to the node is greater than the set data volume proportion, or the preset pruning strategy includes a pruning strategy set according to the attribution analysis business logic.

[0106] If the influence degree is reflected by the contribution degree, then if the contribution degree is greater than the set threshold, it is considered that the influence of this node on the metric change is increased, and at this time, it can be added to the stack for further exploration. On the contrary, if the contribution degree is less than or equal to the set threshold, it is considered that the influence of this node on the metric change is not significant, and at this time, it can be removed from the stack.

[0107] Among them, the data volume proportion may refer to the order volume proportion. For example, if the order volume proportion of a certain node (such as the business type) is small, and the attribution metric to be analyzed is the order timeout rate, and the order volume proportion is small, then its influence on the order timeout rate may not be significant. Therefore, at this time, there is no need to continue exploring the child nodes of this node, and this node can be removed from the stack.

[0108] In another case, the pruning strategy can be set by the user according to the attribution analysis business logic. For example, for the order timeout rate, if the currently explored node is the origin city, and its child node is the origin province, then there is no need to continue exploring the origin province at this time because it is meaningless. Therefore, the pruning strategy also supports user customization according to the business logic.

[0109] In the above implementation method, by setting the pruning strategy, nodes that have little influence on the metric change can be pruned during the search process, which can avoid continuing to deeply explore the child nodes of this node, avoid useless searches, and improve the search efficiency of the algorithm.

[0110] Based on the above embodiments, when applying the pruning strategy, the threshold in the preset pruning strategy can also be determined according to business parameters. This threshold can refer to the set threshold corresponding to the influence degree, or the set data volume proportion corresponding to the data volume proportion, that is, the set threshold or the set data volume proportion can be dynamically adjusted according to business parameters, or when setting the pruning strategy according to the attribution analysis business logic, the influence of business parameters on the business logic is also considered.

[0111] Among them, the business parameters may include parameters such as season, activity type, and holiday.

[0112] Seasons (such as spring, summer, autumn, winter) can affect the performance of certain business metrics. For example, the sales volume of certain products may increase or decrease significantly in specific seasons. In some seasons, a higher threshold may be required for the degree of influence of certain dimensions to be considered significant. For example, in summer, the sales volume of air conditioners may increase significantly. Therefore, when analyzing sales data, the threshold corresponding to the degree of influence of dimensions related to air conditioners or the threshold corresponding to the proportion of data volume can be increased.

[0113] Alternatively, in a specific season, certain dimensions may be more worthy of attention. For example, in winter, the sales of heating equipment may become a dimension of key concern. Therefore, the nodes of these dimensions can be preferentially retained.

[0114] Activity types (such as promotional activities, new product launches, festival activities, etc.) can significantly affect business metrics. For example, promotional activities may lead to large fluctuations in sales. During promotional activities, the degree of influence of certain dimensions may increase significantly. For example, during the "Double Eleven" promotion, the sales volume of certain products may increase significantly. Therefore, the threshold corresponding to the degree of influence of dimensions related to these products or the threshold corresponding to the proportion of data volume can be increased.

[0115] Alternatively, during the new product launch period, the dimension nodes related to the new products can be preferentially retained because these dimensions may have a significant impact on sales data.

[0116] Holidays (such as the Spring Festival, National Day, Christmas, etc.) usually lead to significant changes in business metrics. For example, tourism and consumption activities during the Spring Festival may increase significantly. Before and after holidays, the degree of influence of certain dimensions may increase significantly. For example, during the Spring Festival, the degree of influence of dimensions related to tourism (such as hotel bookings, sales of tourist attraction tickets) may increase significantly. Therefore, the threshold corresponding to the degree of influence of these dimensions or the threshold corresponding to the proportion of data volume can be increased.

[0117] Alternatively, before and after holidays, the dimension nodes related to holidays can be preferentially retained because these dimensions may have a significant impact on business metrics.

[0118] Based on the above embodiments, during the process of traversing the data tree, each time a node is traversed, the corresponding degree of influence is calculated, that is, the influence degree of a single dimension value under the exploration dimension on the overall market index. If the attribution dimension organization form is dimension drilling down, only the first configured dimension is explored, and all dimensions need to be explored for "single dimension" and "combination dimension".

[0119] Taking the order timeout rate as the attribution index to be analyzed, and the degree of influence is characterized by the contribution rate. In the case where the attribution analysis type is dimension achievement attribution, the calculation method of the contribution rate (exclusion method) is as follows:

[0120]

[0121] If the degree of influence is characterized by the degree of surprise, the calculation method of the degree of surprise according to the adtributor algorithm is as follows:

[0122]

[0123]

[0124] If the attribution dimension organization form is configured as a single dimension, the algorithm ends; otherwise, continue with the next operation: take a node from the stack and drill down one level from this node.

[0125] For example, if the obtained node is "indicator name = timeout rate, dimension name = business type, dimension value = brand direct delivery, indicator value = 0.06, contribution degree = 0.002, degree of surprise = 0.001", and at this time it is necessary to continue to explore the dimension "carrier product" below, then the aggregated data can be conveniently obtained from the "data tree": find "dimension name = business type, dimension value = brand direct delivery", and filter out the node with "dimension name = carrier product" from its sub-node list, and then repeat the above operations for the selected node.

[0126] After constructing the single-dimension nodes, start multi-threaded search. Since each time a node is taken from the stack NodeStack for depth-first search, there is no mutual influence. Search layer by layer until NodeStack is empty, indicating that all nodes have been explored.

[0127] In the case where the attribution analysis type is dimension fluctuation attribution, the process is similar to the above dimension achievement attribution, and the differences are as follows: There are two data trees constructed, that is, the nodes to be obtained each time are the data of the same dimension in two cycles. It should be noted that it is possible that the data in the two cycles cannot be completely corresponding, that is, there may be no data for the dimension value under a certain dimension name in one cycle. At this time, it can be filled with 0. The reason for not directly discarding this part is that these nodes are very likely to belong to the root cause because the change is large, from none to some or from some to none.

[0128] The calculation method of the contribution degree is also different. In this method, the fixed-base method can be used. Replace the data under the dimension combination in the comparison period with the data under the same dimension combination in the current period, and then subtract the indicator value in the comparison period. The calculation process is as follows:

[0129]

[0130] On the basis of the above embodiments, the attribution analysis type may further include factor fluctuation attribution. Different from dimension fluctuation attribution, dimension fluctuation attribution decomposes the fluctuation of the indicator value into each dimension to analyze which dimensions are the root causes of the abnormal fluctuation, while factor fluctuation attribution decomposes this fluctuation into each factor of the indicator calculation formula. For example, the calculation formula of an indicator is (A - B) / C, which contains three factors A, B, and C. In two periods, it can be analyzed which factor's change brings a large fluctuation to the indicator.

[0131] When the attribution analysis type included in the relevant indicator configuration parameters is factor fluctuation attribution, during the analysis, the type of the indicator calculation formula corresponding to the attribution indicator to be analyzed can be determined first, and then the calculation method of the influence degree can be determined according to the type of the indicator calculation formula. According to the corresponding calculation method, the influence degree of each calculation factor in the indicator calculation formula on the attribution indicator to be analyzed is calculated.

[0132] Among them, the types of the indicator calculation formula can include ratio type, multiplication type, and addition type. For the ratio type, its calculation method is the fixed-base method; for the multiplication type, its calculation method is the LMDI (Log-Mean Divisia Index) algorithm; for the addition type, its calculation method is the indicator decomposition method.

[0133] Among them, the calculation process of the fixed-base method is as follows:

[0134] Select a divisor or dividend in the indicator calculation formula;

[0135] The results of the indicators corresponding to each dimension in the previous and later periods are respectively

[0136] The target indicator is used as the Baseline base period, where: the previous value is The later value is

[0137] The contribution degree calculation formula of the divisor factor is:

[0138]

[0139] Normalize all the contribution degrees in this group (divide by the sum of the contribution degrees) to obtain the normalized contribution degree

[0140] Multiply the normalized contribution degree after normalizing the contribution degree by the change value of the target factor before and after to obtain the contribution value CVs i = CRs i ·(X2 - X1).

[0141] The calculation process using LMDI is as follows:

[0142] Decompose the multiplier factor in the index calculation formula: FR t = D t × S t × P t × B t × 100%;

[0143] Take the logarithms Ln(S1) and Ln(S2) of the results of each multiplier factor (D / S / P / B) for the previous and current periods;

[0144] Calculate the difference in logarithms of each multiplier factor, Ln(S2) - Ln(S1);

[0145] Divide the difference in logarithms of each multiplier factor by the difference in logarithms of the target index to calculate the contribution degree

[0146] Sort the contribution degrees CR from high to low: Rank(1, 2, 3, 4) = B, P, S, D;

[0147] Contribution degree CV = factor contribution degree * (target index comparison value - base period value).

[0148] The calculation process of the index decomposition method is as follows:

[0149] For example, taking the abnormal arrival quantity in the seller's fulfillment conversion rate factor as an example by business type dimension, assume the following:

[0150] X is the abnormal arrival quantity, X1, X2;

[0151] Xa, Xb, Xc are the abnormal arrival order quantities of spot goods, consignment, and brand direct shipments respectively;

[0152] Xa1, Xa2 are the previous and current values of the abnormal arrival quantity of spot goods respectively;

[0153] Verify whether all are fully decomposed: X1 = Xa1 + Xb1 + Xc1;

[0154] Verification method: On the premise of meeting the MECE principle, ensure that there are no omissions in the dimension splitting addition factors;

[0155] Taking spot goods as an example, calculate the volatility contribution degree CR a = (Xa2 - Xa1) ÷ X1;

[0156] Normalize the contribution degrees of all addition factors within the dimension (divide by the sum of the contribution degrees) to obtain the normalized contribution degree:

[0157]

[0158] In the above implementation process, for factor fluctuation attribution, different calculation methods are selected according to the type of calculation formula to calculate the impact degree, which can improve the accuracy of calculating the impact degree.

[0159] Based on the above embodiments, the sum of the contribution degrees of all dimensions may not be 1 because of pruning. For the convenience of understanding and comparison, the contribution degrees can be normalized during post-processing so that the sum of the positive and negative contribution degrees is 1. Specifically, the impact degree can be normalized according to the index attributes of the attribution index to be analyzed, and the impact degree after normalization is obtained, where the index attributes include positive attributes and negative attributes.

[0160] In actual business, indicators are divided into positive and negative. For example, the overtime rate is a negative indicator (the lower the better), and the on-time delivery rate for outbound is a positive indicator (the higher the better). The positive and negative of the contribution degree itself refer to the positive and negative impacts on the indicator anomaly. Therefore, the two directions need to be considered together. For example, the positive contribution degree of the negative indicator overtime rate is to increase the overtime rate, which should be a negative impact. Intuitively, the understanding of the contribution degree is that positive contribution is a positive impact and negative contribution is a negative impact. Therefore, a normalization conversion is performed during post-processing, and the conversion logic is shown in the following table:

[0161]

[0162] In the above implementation process, the normalization process can clarify the contribution direction (positive or negative) of each dimension or factor to the change of the indicator, so that users can quickly know the root causes with greater influence through the attribution analysis results.

[0163] Please refer to Figure 5 , Figure 5 FIG. 200 is a structural block diagram of an attribution analysis device 200 provided by an embodiment of the present application. The device 200 may be a module, a program segment, or code on an electronic device. It should be understood that the device 200 corresponds to the above Figure 1 method embodiment and can execute Figure 1 each step involved in the method embodiment. The specific functions of the device 200 can be seen in the above description. To avoid repetition, the detailed description is appropriately omitted here.

[0164] Optionally, the device 200 includes:

[0165] A data acquisition module 210, configured to determine an attribution index to be analyzed and related index configuration parameters. The related index configuration parameters include the attribution analysis type configured by the user, the attribution dimension organization form, and the attribution dimension. The attribution analysis type includes dimension achievement attribution and dimension fluctuation attribution. The attribution dimension organization form includes single dimension, combined dimension, and dimension drilling down;

[0166] A calculation module 220, configured to calculate the influence degree of each attribution dimension on the attribution index to be analyzed according to the parameters configured by the relevant indicators;

[0167] A result determination module 230, configured to determine an attribution analysis result according to the influence degree.

[0168] Optionally, the calculation module 220 is configured to traverse a constructed data tree according to the attribution dimensions in the parameters configured by the relevant indicators, where each node in the data tree includes a dimension name, a dimension value, an index value, and a child node of the attribution dimension; for each traversed node, calculate the influence degree of the corresponding attribution dimension of the node on the attribution index to be analyzed.

[0169] Optionally, the data tree is constructed in the following manner:

[0170] Obtain the original data to be analyzed from a data source according to the attribution index to be analyzed and the attribution dimensions in the parameters configured by the relevant indicators, where the original data includes multiple attribution dimensions and corresponding index values;

[0171] Perform aggregation processing on the original data according to a preset dimension combination rule to generate aggregation data at different levels, and each level of aggregation data corresponds to a combination of one or more attribution dimensions;

[0172] Generate a hierarchical data tree according to the aggregation data at different levels.

[0173] Optionally, the parameters configured by the relevant indicators further include a data date range configured by a user, and the calculation module 220 is configured to obtain the original data to be analyzed from a data source according to the attribution index to be analyzed, the attribution dimensions in the parameters configured by the relevant indicators, and the data date range.

[0174] Optionally, the calculation module 220 is configured to traverse the constructed data tree by using a depth-first search algorithm according to the attribution dimensions in the parameters configured by the relevant indicators.

[0175] Optionally, the calculation module 220 is further configured to prune the traversed nodes according to a preset pruning strategy, and put the nodes that meet the preset pruning strategy into a stack as candidate search nodes for the subsequent depth-first search algorithm.

[0176] Optionally, the preset pruning strategy includes that the influence degree is greater than a set threshold, or the preset pruning strategy includes that the data volume ratio corresponding to the node is greater than a set data volume ratio, or the preset pruning strategy includes a pruning strategy set according to the business logic of attribution analysis.

[0177] Optionally, the attribution analysis type further includes factor fluctuation attribution. When the attribution analysis type included in the correlation index configuration parameter is factor fluctuation attribution, the result determination module 230 is further configured to determine the type of the index calculation formula corresponding to the attribution index to be analyzed; determine the calculation method of the influence degree according to the type of the index calculation formula; calculate the influence degree of each calculation factor in the index calculation formula on the attribution index to be analyzed according to the calculation method; and determine the attribution analysis result according to the influence degree.

[0178] Optionally, the apparatus 200 further includes:

[0179] A normalization module, configured to perform normalization processing on the influence degree according to the index attribute of the attribution index to be analyzed, so as to obtain the normalized influence degree, where the index attribute includes a positive attribute and a negative attribute.

[0180] It should be noted that those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the above-described system can refer to the corresponding process in the foregoing method embodiment, and will not be described herein again.

[0181] Please refer to Figure 6 , Figure 6 FIG. for a schematic structural diagram of an electronic device for executing an attribution analysis method provided by an embodiment of the present application. The electronic device may include: at least one processor 310, such as a CPU, at least one communication interface 320, at least one memory 330, and at least one communication bus 340. Among them, the communication bus 340 is used to implement connection communication between these components. Among them, the communication interface 320 of the device in the embodiment of the present application is used to communicate with other node devices for signaling or data. The memory 330 may be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The memory 330 may optionally be at least one storage device located far from the foregoing processor. The memory 330 stores computer-readable instructions. When the computer-readable instructions are executed by the processor 310, the electronic device executes the above Figure 1 shown method process.

[0182] It can be understood that Figure 6 The structure shown is only schematic, and the electronic device may further include more or fewer components than those shown in Figure 6 , or have a different configuration from that shown in Figure 6 . Figure 6 Each component shown in can be implemented by hardware, software, or a combination thereof.

[0183] An embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it executes the method process performed by the electronic device in the method embodiment as shown in Figure 1 The method process performed by the electronic device in the method embodiment as shown.

[0184] This embodiment discloses a computer program product. The computer program product includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the methods provided in the above method embodiments. For example, it includes:

[0185] Determine the attribution index to be analyzed and the relevant index configuration parameters. The relevant index configuration parameters include the attribution analysis type configured by the user, the attribution dimension organization form, and the attribution dimension. The attribution analysis type includes dimension achievement attribution and dimension fluctuation attribution. The attribution dimension organization form includes single dimension, combined dimension, and dimension drill-down;

[0186] According to the relevant index configuration parameters, calculate the influence degree of each attribution dimension on the attribution index to be analyzed;

[0187] According to the influence degree, determine the attribution analysis result.

[0188] In summary, the embodiments of the present application provide an attribution analysis method, device, electronic device, storage medium, and program product. By allowing the user to configure the relevant index configuration parameters and automatically implementing the analysis through the system, the efficiency and accuracy are higher. Moreover, it supports attribution analysis of different attribution analysis types and different attribution dimension organization forms, with higher flexibility and can meet the different analysis needs of users.

[0189] In the embodiments provided by the present application, it should be understood that the disclosed device and method can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For another example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling, direct coupling, or communication connection can be through some communication interfaces. The indirect coupling or communication connection of the device or unit can be in an electrical, mechanical, or other form.

[0190] In addition, the units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0191] Furthermore, in each embodiment of the present application, each functional module may be integrated together to form an independent part, or each module may exist alone, or two or more modules may be integrated to form an independent part.

[0192] In this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations.

[0193] The above description is only for the embodiments of the present application and is not intended to limit the protection scope of the present application. For those skilled in the art, the present application may have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An attribution analysis method, characterized in that, The method includes: Determine the attribution metrics to be analyzed and the relevant metric configuration parameters. The relevant metric configuration parameters include the attribution analysis type configured by the user, the attribution dimension organization form, and the attribution dimensions. The attribution analysis type includes dimension achievement attribution and dimension fluctuation attribution. The attribution dimension organization form includes single dimension, combined dimension, and dimension drill-down; Calculate the influence degree of each attribution dimension on the attribution metrics to be analyzed according to the relevant metric configuration parameters; Determine the attribution analysis result according to the influence degree.

2. The method according to claim 1, wherein The calculating the influence degree of each attribution dimension on the attribution metrics to be analyzed according to the relevant metric configuration parameters includes: Traverse the constructed data tree according to the attribution dimensions in the relevant metric configuration parameters. Each node in the data tree includes the dimension name, dimension value, metric value, and sub-nodes of the attribution dimension; For each traversed node, calculate the influence degree of the corresponding attribution dimension of this node on the attribution metrics to be analyzed.

3. The method according to claim 2, wherein The data tree is constructed in the following manner: Obtain the original data to be analyzed from the data source according to the attribution metrics to be analyzed and the attribution dimensions in the relevant metric configuration parameters. The original data includes multiple attribution dimensions and corresponding metric values; Perform aggregation processing on the original data according to the preset dimension combination rules to generate aggregation data at different levels. Each level of aggregation data corresponds to a combination of one or more attribution dimensions; Generate a hierarchical data tree according to the aggregation data at different levels.

4. The method according to claim 3, characterized in that, The relevant metric configuration parameters further include the data date range configured by the user. The obtaining the original data to be analyzed from the data source according to the attribution metrics to be analyzed and the attribution dimensions in the relevant metric configuration parameters includes: Obtain the original data to be analyzed from the data source according to the attribution metrics to be analyzed, the attribution dimensions in the relevant metric configuration parameters, and the data date range.

5. The method according to claim 2, characterized in that, The traversing the constructed data tree according to the attribution dimensions in the relevant metric configuration parameters includes: Traverse the constructed data tree using the depth-first search algorithm according to the attribution dimensions in the relevant metric configuration parameters.

6. The method according to claim 5, characterized in that, After calculating the influence degree of the corresponding attribution dimension of each traversed node on the attribution metrics to be analyzed, it further includes: Prune the traversed nodes according to the preset pruning strategy, and put the nodes that meet the preset pruning strategy into the stack as candidate search nodes for the subsequent depth-first search algorithm.

7. The method according to claim 6, characterized in that, The preset pruning strategy includes that the influence degree is greater than the set threshold, or the preset pruning strategy includes that the data volume ratio corresponding to the node is greater than the set data volume ratio, or the preset pruning strategy includes a pruning strategy set according to the attribution analysis business logic.

8. The method according to claim 1, wherein The attribution analysis type further includes factor fluctuation attribution. When the attribution analysis type included in the relevant metric configuration parameters is factor fluctuation attribution, the method further includes: Determine the type of the metric calculation formula corresponding to the attribution metrics to be analyzed; Determine the calculation method of the influence degree according to the type of the metric calculation formula; According to the described calculation method, calculate the influence degree of each calculation factor in the index calculation formula on the attribution index to be analyzed; Determine the attribution analysis result according to the influence degree.

9. The method according to any one of claims 1-8, characterized in that, After obtaining the influence degree, it further includes: According to the index attribute of the attribution index to be analyzed, perform normalization processing on the influence degree to obtain the normalized influence degree, where the index attribute includes a positive attribute and a negative attribute.

10. An attribution analysis device, characterized in that, The device includes: A data acquisition module, configured to determine an attribution index to be analyzed and related index configuration parameters, where the related index configuration parameters include the attribution analysis type configured by the user, the attribution dimension organization form, and the attribution dimension. The attribution analysis type includes dimension achievement attribution and dimension fluctuation attribution, and the attribution dimension organization form includes a single dimension, a combined dimension, and dimension drilling down; A calculation module, configured to calculate the influence degree of each attribution dimension on the attribution index to be analyzed according to the related index configuration parameters; A result determination module, configured to determine the attribution analysis result according to the influence degree.

11. An electronic device, characterized in that, It includes a processor and a memory, and the memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the method according to any one of claims 1-9 is run.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the method according to any one of claims 1-9 is run.

13. A computer program product, characterized in that, It includes computer program instructions. When the computer program instructions are read and run by the processor, the method according to any one of claims 1-9 is executed.

Images