An automatic mining method of runtime requirements based on user operation sequence

Abstract

The application discloses a method for automatically mining runtime requirements based on user operation sequences and target models of an adaptive website platform, which comprises the following three main steps: step (1) associating target models with operation sequences; step (2) mining behavior patterns; and step (3) generating runtime target models; the application proposes an innovative method for the requirement engineering field of the adaptive website platform. The method generates runtime target models based on user operation sequences and target models of the adaptive website platform, thereby helping analysts of the platform to obtain runtime requirements. Therefore, the application can reduce the work burden of the analysts, improve work efficiency, and provide stronger support for the design and optimization of the adaptive website platform.

Description

Technical Field

[0001] This invention applies to the field of requirements engineering for adaptive website platforms, proposing a method for automatically mining runtime requirements based on user operation sequences. This method involves goal-oriented requirements modeling and data mining techniques, using automated methods to accurately obtain runtime requirements based on operation sequences. Background Technology

[0002] Adaptive systems are software systems with intelligent adjustment and optimization capabilities, automatically adapting their behavior and performance to different environmental conditions and user needs. Adaptive website platforms, for example, adjust their layout and functionality to provide a consistent and positive user experience. To achieve this adaptability, website platforms need to acquire and understand runtime requirements in a timely manner to make corresponding adjustments and optimizations. However, most current adaptive website platforms primarily focus on runtime environmental data, such as physical parameters like device size and lighting, when acquiring runtime requirements. While this environmental data is crucial for website platform adaptability, it overlooks another critical factor: user runtime behavior. User runtime behavior information provides vital information about system usage patterns, preferences, and expectations, thus more accurately reflecting user needs and intentions.

[0003] Traditional goal models typically only consider static design-time requirements when describing system needs, lacking sufficient consideration of runtime behavior. To address this deficiency, researchers have proposed the Runtime Goal Model (RGM), which supplements traditional goal models with runtime behavioral information, providing a more comprehensive representation of runtime requirements. Through the RGM, website platforms can better understand user behavior and needs, thus making appropriate adaptive adjustments. However, currently, building RGMs primarily relies on manual methods, requiring professional analysts to spend significant time and effort collecting and organizing runtime behavioral information. This manual approach suffers from inefficiency, error-proneness, and limited scalability.

[0004] To address the above problems, this invention proposes an innovative method for automatically mining the runtime requirements of adaptive website platforms. This method utilizes goal-oriented requirement modeling and data mining techniques to combine user operation sequences of the website platform with a target model, automatically constructing a runtime target model to obtain runtime requirements. This invention has broad application prospects in adaptive website platform optimization, user behavior analysis, and requirement mining, providing important technical support for the development and application of adaptive systems. Summary of the Invention

[0005] This invention proposes a method for automatically mining platform runtime requirements based on user operation sequences and target models of an adaptive website platform. This invention mainly includes the following three steps:

[0006] Step (1) Associate the target model with the operation sequence;

[0007] To analyze the user behavior information of each target node in the target model of the website platform, this step requires accurately associating the target nodes in the target model with the corresponding segments in the user operation sequence of the website platform, obtaining a set of subsequences corresponding to all target nodes. This process involves two sub-steps.

[0008] Step (1.1) Filter the operation sequence;

[0009] To ensure that suitable sequences are obtained for analyzing user behavior information, operation sequences that do not provide sufficient runtime information are filtered out.

[0010] Step (1.2) matches the subsequence corresponding to the target node;

[0011] After obtaining valuable operation sequences in step 1.1, corresponding subsequences are extracted for target nodes in the target model based on the correspondence between the target model and the user operation sequences. One target node typically corresponds to multiple subsequences. The subsequence set consists of all subsequences corresponding to all target nodes in the target model.

[0012] Step (2) Discover behavioral patterns;

[0013] Based on the set of subsequences obtained in step 1, multi-level behavioral patterns are mined for each target node. Multi-level behavioral patterns are behavioral expressions at different levels of abstraction, which can describe the user's runtime behavioral information. This process involves five sub-steps.

[0014] Step (2.1) Customize the mining dimensions;

[0015] This method considers only one basic mining dimension by default: the operations appearing in the sequence. Therefore, analysts are allowed to customize and select other additional analysis dimensions according to their needs.

[0016] Step (2.2) Extract the common prefix;

[0017] After determining the mining dimensions in step 2.1, the common prefix of all subsequences corresponding to the target node is extracted. A common prefix refers to the common operation at the starting position of most subsequences. This step recursively extracts common prefixes until none are found.

[0018] Step (2.3) summarizes the remaining parts;

[0019] Further processing is performed on the expression after extracting the common prefix in step 2.2. The remaining part after extracting the common prefix is ​​summarized to simplify the expression and make it easier to understand.

[0020] Step (2.4) Identify repetitive operations;

[0021] The expression obtained in step 2.3 is further processed to identify repeated operations in the expression, and the expression is further simplified to eliminate repetition and make it more concise.

[0022] Step (2.5) Supplement the sequence information;

[0023] The simplified expression from step 2.4 is further processed to supplement the expression with the sequential relationship between operations, so as to more accurately describe the user's runtime behavior information.

[0024] Step (3) Generate the runtime target model;

[0025] After obtaining the multi-level behavioral patterns of each target node in step 2, the analysts of the adaptive website platform select the most valuable behavioral patterns. Based on the analysts' selections, we generate a runtime target model, which provides runtime behavioral information of users. This allows us to analyze users' runtime needs when using the website, providing an important basis for optimizing the adaptive website platform.

[0026] The beneficial effects of this invention are: it proposes an innovative method for requirements engineering of adaptive website platforms. This method generates a runtime target model based on the user operation sequence and target model of the website platform, thereby helping platform analysts obtain runtime requirements. Therefore, this invention can reduce the workload of analysts, improve work efficiency, and provide stronger support for the design and optimization of adaptive website platforms. Attached Figure Description

[0027] Figure 1 This is a flowchart of the method for automatically mining runtime requirements proposed in this invention.

[0028] Figure 2 This summarizes the remaining algorithms in this invention.

[0029] Figure 3 An example of the target model used as input for this invention.

[0030] Figure 4 This is an example of the runtime target model generated by the present invention. Detailed Implementation

[0031] A brief overview of the invention is given below to provide a basic understanding of certain aspects of it. It should be understood that this overview is not an exhaustive summary of the invention. It is not intended to identify key or essential parts of the invention, nor is it intended to limit the scope of the invention. Its purpose is merely to present certain concepts in a simplified form as a prelude to the more detailed description that follows.

[0032] The flowchart of the method for automatically mining runtime requirements of adaptive website platforms based on user operation sequences and target models proposed in this invention is as follows: Figure 1 As shown, the method consists of three main steps: step (1) associating the target model with the operation sequence, step (2) mining behavioral patterns, and step (3) generating the runtime target model (RGM). The following describes each step in detail with reference to the accompanying drawings and examples.

[0033] Step (1) Associate the target model with the operation sequence;

[0034] The user's action sequence during website usage can be obtained from the backend data of the adaptive website platform, while the website designer can provide the target model during design. Both serve as inputs to this invention. An example of the target model during design is shown below. Figure 3 As shown. In this step, the target model is associated with the operation sequence to obtain a set of subsequences.

[0035] Step (1.1) Filter the operation sequence;

[0036] A user action sequence refers to a series of actions within a complete user session, that is, all actions taken by the user from the start of using the website platform to completing a specific goal or exiting the platform. A short user action sequence cannot contain enough relevant runtime information to describe user behavior. Therefore, a length threshold of 10 is set in this step to filter sequences. The length of each action sequence is checked individually, and any sequence shorter than this threshold is discarded.

[0037] Step (1.2) matches the subsequence corresponding to the target node;

[0038] After obtaining the valuable operation sequence in step 1.1, the subsequences corresponding to the target nodes are matched according to the correspondence. In this invention, the target model input conforms to the ISAR 2.0 standard. According to this standard, leaf nodes represent user operations during the use of the adaptive website platform, while other non-leaf nodes (target nodes) represent the user's expected results. Since the leaf nodes in the target model correspond to the operations in the user's operation sequence, this step uses regular expressions to extract the corresponding subsequences of each target node in the target model from bottom to top based on the correspondence. Typically, users complete various goals in an interleaved manner during the use of the website platform; therefore, a single goal usually has multiple corresponding subsequences. The subsequence set output in this step consists of all subsequences corresponding to all target nodes.

[0039] Step (2) Discover behavioral patterns;

[0040] After obtaining the subsequences corresponding to each target node in step 1, the behavioral information expressed in the sequences is mined to obtain the multi-level behavioral patterns of each target node. This mining process will generate five types of expression information, as shown in Table 1.

[0041] Table 1 Expression Information

[0042]

[0043] A subsequence corresponding to a target node represents one type of user behavior during the use of the website platform. Therefore, we first use the "|" relational symbol in Table 1 to concatenate all subsequences corresponding to a target node, in order to comprehensively analyze the general characteristics of the target node's runtime behavior during the use of the adaptive website platform.

[0044] Step (2.1) Customize the mining dimensions;

[0045] Before mining behavioral patterns, the mining dimensions must be determined. This invention defaults to considering only one basic analytical dimension: the operations appearing in the sequence. Analysts can customize the analysis to include other additional dimensions as needed, such as the time interval between operations or the frequency of operations. Therefore, this step first determines whether the analyst has requested any additional analytical dimensions. If so, additional information is added based on the requested new dimensions.

[0046] Step (2.2) Extract the common prefix;

[0047] After determining the mining dimensions in step 2.1, the common prefix of the subsequences corresponding to each target node is extracted. First, the threshold is determined according to the proposed formula.

[0048]

[0049] Traverse the subsequences, recording the frequency of each operation at the beginning of the subsequence. If the frequency is greater than or equal to a threshold, it is considered a common prefix. If a common prefix exists, extract it, and the representation will change from subsequence 1|subsequence 2|... to common prefix (remaining part 1|remaining part 2|...). If the remaining part is empty, it is marked as skip in Table 1. Next, repeat the above steps to extract the common prefixes at lower levels of abstraction in the remaining part until no common prefix is ​​found (no common prefix exists or the remaining part is completely empty).

[0050] Each step in the process of extracting common prefixes deals with expressions at different levels of abstraction. Therefore, by saving the results of each step, we can obtain multi-level expressions.

[0051] Step (2.3) summarizes the remaining parts;

[0052] After extracting the common prefix in step 2.2, the remaining part of the expression does not contain general features and therefore cannot effectively explain the user's runtime operations. Therefore, an algorithm for summarizing the remaining part is designed. This algorithm simplifies the expression by summarizing the operations contained within the remaining part. The algorithm is as follows: Figure 2 As shown.

[0053] Step (2.4) Identify repetitive operations;

[0054] After summarizing the remaining part of the expression in step 2.3, the simplified expression is traversed to identify repeated operations. If a part of the sequence consists of a single operation repeated two or more times, that part is replaced with the form "O+", where O represents the single operation.

[0055] Step (2.5) Supplement the sequence information;

[0056] Iterate through the expressions obtained in step 2.4, supplementing them with information indicating the order of operations within the expressions. When there is an order relationship between two operations, we insert a semicolon (";") from Table 1 between them to indicate the order relationship between the two operations.

[0057] After completing the above 5 sub-steps, a multi-level behavior pattern can be obtained in the form of an expression based on all the sub-sequences corresponding to the target node.

[0058] Step (3) Generate the runtime target model;

[0059] After obtaining the multi-level behavioral patterns of each target node in step 2, the analysts of the adaptive website platform select the behavioral patterns they deem most valuable. Next, the selected behavioral patterns are added to the corresponding target nodes in the target model, supplementing the runtime behavioral information of that node and generating the runtime target model. As a form of expression of runtime requirements, the runtime target model effectively assists analysts in obtaining runtime behavioral information of users using the adaptive website platform, thereby obtaining valuable runtime requirements. An example of a runtime target model is shown below. Figure 4 As shown.

[0060] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for automatically mining runtime requirements based on user operation sequences and target models of an adaptive website platform, characterized in that: This method consists of three steps: Step (1) Associate the target model with the operation sequence; obtain the user's operation sequence during the use of the website from the backend data of the adaptive website platform. The website designer provides the target model during the design process. The user operation sequence and the target model are used as inputs; according to the correspondence, associate the target model with the operation sequence to obtain a set of subsequences; Step (2) Mining behavioral patterns; After obtaining the subsequences corresponding to each target node in step (1), mining the behavioral information expressed in the sequence to obtain the multi-level behavioral patterns of each target node; A subsequence corresponding to a target node is a situation of user behavior during the use of the website platform; First, use the "|" relation symbol to concatenate all the subsequences corresponding to a target to comprehensively analyze the general characteristics of the target node's runtime behavior during the user's use of the adaptive website platform; Step (3) Generate the runtime target model (RGM); After obtaining the multi-level behavior patterns of each target node in step (2), the analysts of the adaptive platform website select the most valuable behavior pattern; Add the selected behavior pattern to the corresponding target node in the target model to supplement the runtime behavior information of the node and generate the runtime target model; The runtime target model, as a form of expression of runtime requirements, assists analysts in obtaining runtime behavior information of users in the process of using the adaptive website platform, thereby obtaining valuable runtime requirements and providing strong support for the design and optimization of the adaptive website platform.

2. The method for automatically mining runtime requirements based on user operation sequences and target models of an adaptive website platform according to claim 1, characterized in that: Step (1) Associate the target model with the operation sequence; Step (1.1) Filter the operation sequence; Set a length threshold to filter sequences; check the length of each sequence in the operation, and discard any sequence that is shorter than this length threshold; Step (1.2) matches the subsequence corresponding to the target node; After obtaining the operation sequence with analytical value through step (1.1), the corresponding subsequence of each target node in the target model is extracted from the bottom up using regular expressions based on the correspondence; the output is a set of subsequences consisting of all subsequences corresponding to all target nodes. Step (2) Discover behavioral patterns; Step (2.1) Customize the mining dimensions; Before mining behavioral patterns, first determine the mining dimensions; by default, only one basic analysis dimension is considered, namely the operations that appear in the sequence; analysts of adaptive website platforms can customize other additional analysis dimensions as needed; first determine whether the analyst has requested any additional analysis dimensions; if so, add additional information according to the requested new dimensions; Step (2.2) Extract the common prefix; After determining the mining dimensions in step (2.1), the common prefix of the subsequences corresponding to each target node is extracted; first, the threshold is determined according to the proposed formula; Traverse the subsequence and record the number of times each operation appears at the beginning of the subsequence; if the number of occurrences is greater than or equal to the threshold, it is considered a common prefix; if a common prefix exists, extract it, and if the remaining part is empty, mark it as "skip"; repeat the above steps to extract the common prefixes of the lower level of abstraction of the remaining part until no common prefix can be found. Each step in the process of extracting common prefixes deals with expressions at different levels of abstraction. By saving the results of each step, we can obtain multi-level expressions. Step (2.3) summarizes the remaining parts; After extracting the common prefix in step (2.2), the remaining part of the expression does not contain general features, and therefore cannot effectively explain the user's runtime operation information; an algorithm is implemented to summarize the remaining part based on the operation information it contains to simplify the expression; Step (2.4) Identify repetitive operations; After summarizing the remaining part of the expression in step 2.3, the simplified expression is traversed to identify repeated operations; if a part of the sequence consists of a single operation repeated two or more times, the part is replaced with "O+", where O represents the single operation. Step (2.5) Supplement the sequence information; Traverse the expression obtained in step (2.4) and supplement it with information indicating the order of operations in the expression; when there is an order relationship between two operations, a ";" symbol will be inserted between them to indicate the order relationship between the two operations; After completing the above five sub-steps (2.1) to (2.5), the multi-level behavior pattern corresponding to the node is obtained based on all the sub-sequences corresponding to the target node. This behavior pattern is presented in the form of an expression. Step (3) Generate the runtime target model.

3. The method for automatically mining runtime requirements based on user operation sequences and target models of an adaptive website platform according to claim 2, characterized in that: Mining behavioral patterns generates five types of expression information: skip is empty; E+ indicates one or more sequential occurrences; E* indicates zero or more sequential occurrences; E1|E2 indicates one selection; E1;E2 indicates sequential occurrences.

Images