Business process control method and system based on decision engine

Abstract

The invention discloses a business process control method and system based on a decision engine, and relates to the technical field of business process control. The method comprises the following steps: collecting environmental change data in a business process in real time, classifying and filtering the data, and dividing the data into a significant change group and a tiny fluctuation group to obtain an environmental change core set; strategy matching analysis is carried out on the significant change group, and an adaptive decision adjustment scheme set is determined; then, the execution efficiency and stability of all the schemes are evaluated through historical execution records, comprehensive sorting is carried out, and candidate decision combinations are determined; and finally, evaluating the applicability of the candidate scheme in combination with environmental characteristics. According to the method, quick response to environment change is realized, through historical data analysis and multi-dimensional evaluation, the decision making process is optimized, the decision making accuracy and execution efficiency are improved, the applicability of a complex and changeable business environment is improved, and dynamic optimization and continuous improvement of decision making are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of business process control, and in particular to a business process control method and system based on a decision engine. Background Art

[0002] In modern enterprise management, intelligent control of business processes is a key research direction for achieving operational efficiency and maximizing resource utilization. With the rapid development of information technology, business process control methods based on decision engines have become a crucial means of improving decision quality and response speed. Their importance is self-evident, as they can help companies achieve dynamic adaptation and precise management, especially in complex and volatile market environments.

[0003] However, many current business process control methods suffer from profound deficiencies in practical application: 1. Over-reliance on real-time data: This lacks consideration for decision stability, leading to frequent policy adjustments in the face of subtle environmental changes, resulting in wasted resources and reduced execution efficiency. 2. Ignoring historical execution performance analysis: When addressing decision continuity, existing solutions fail to comprehensively analyze historical execution performance, making it difficult to find a balance in a dynamic environment.

[0004] One of the core challenges facing existing technologies is maintaining decision-making stability amidst environmental fluctuations. This is specifically addressed in two areas: 1. Avoiding unnecessary strategy switching: Decision engines must manage their sensitivity to environmental changes, effectively filtering out minor disruptions to prevent the system from becoming mired in frequent adjustments that impact business process continuity. 2. How to reasonably determine the duration of decision-making: Historical decision-making performance data must be properly evaluated and utilized to avoid execution biases caused by blindly extending or shortening decision cycles.

[0005] Therefore, designing a mechanism that can suppress the interference of small environmental fluctuations on decision-making and reasonably determine the duration of decision-making based on historical effect evaluation is a key issue that needs to be urgently addressed. Summary of the Invention

[0006] The present invention provides a method and system for controlling a business process based on a decision engine to solve the above-mentioned technical problems.

[0007] For this reason, the technical solution of the present invention is achieved as follows: In a preferred embodiment, a method for controlling a business process based on a decision engine includes the following steps: S1: Collects environmental change data in business processes in real time and classifies it into significant change group and minor fluctuation group according to the size of the environmental change; S2: Compare the data of the small fluctuation group with the preset threshold and filter it. If it is lower than the preset threshold, it is marked as a data group that does not need to be adjusted, and the filtered environmental change core set is obtained; S3: After obtaining the data features of the significant change group from the core set of environmental changes, the significant change group is then subjected to strategy matching analysis using a decision optimization algorithm to generate a set of decision adjustment solutions; S4: Use the historical analysis model to evaluate the decision adjustment plan set, compare it with the historical decision effect data, and obtain the execution efficiency value and stability index of each plan; S5: Based on the execution efficiency value and stability index, the decision adjustment scheme set is comprehensively ranked, and the scheme with a stability index higher than the preset benchmark value is selected as the candidate decision combination; S6: Combine the historical decision effect data corresponding to the candidate decision combination with the core set characteristics of the environmental change to obtain the applicability evaluation results of the candidate decision combination in the current environment.

[0008] In the preferred solution, S1 specifically includes: Collect raw environmental data sets containing temperature, humidity, and vibration parameters in real time through a sensor network; The K-means clustering algorithm was used for classification processing to divide the data into a significant change group and a small fluctuation group, and the classified environmental change data set was obtained; According to the classified environmental change data set, the feature vectors of the significant change group are extracted to generate the significant change feature set; Principal component analysis is performed on the significant change feature set, and after dimensionality reduction processing, it is compared with the preset threshold. If it is lower than the preset threshold, high-correlation features are screened through feature selection algorithm, and then time series analysis method is used to extract environmental change trends.

[0009] In the preferred solution, S2 specifically includes: A threshold judgment is performed on the small fluctuation group. If the data value in a certain data group is lower than the preset threshold, it is marked as not requiring adjustment and eliminated to obtain the core set of environmental changes; The support vector machine algorithm is used to identify key change points in the core concentration of environmental changes, and change points that exceed the preset range are classified as key monitoring data groups; Obtain the distribution characteristics of key monitoring data groups, use information entropy calculation to determine the correlation between changes in each group, and obtain the final monitoring priority ranking; Real-time data collection and update processing are carried out for high-priority data groups to obtain the latest dynamic data sets of environmental changes.

[0010] In the preferred solution, S3 specifically includes: After obtaining preliminary structural information of the core data set and determining the scope and categories of significant changes, feature extraction is performed and key indicators are quantified to obtain a feature vector set. This is then input into a pre-built decision optimization model for analysis. If the magnitude of the feature vector change exceeds a preset threshold, the strategy matching process is triggered to determine the appropriate strategy direction. Then, based on the results of the strategy matching process, the decision adjustment content corresponding to the change group is obtained, and the adjustment content is grouped using classification technology to determine the decision adjustment plan set, where: If the adjustments after grouping are consistent with the trend of environmental changes, the plan will be further optimized; Obtain verification data related to the current adaptation plan, compare it with the optimized plan, judge the feasibility of the plan set, and obtain the final decision adjustment plan set.

[0011] In the preferred solution, the S4 specifically includes: Extract historical decision-making effect data from historical execution records, perform data cleaning, remove outliers and missing values, and obtain a standardized decision-making effect data set; Construct a feature vector including execution time, resource consumption and result quality to obtain the solution feature set; The random forest algorithm is used to classify the feature set of the schemes to obtain the execution efficiency value and stability index of each scheme.

[0012] In a preferred embodiment, the step S5 specifically includes: If the stability index is higher than the preset threshold, the corresponding solution will be marked as a priority; The marked decision adjustment plan set is comprehensively ranked using the stability index combined with the execution efficiency value through a weighted scoring algorithm; According to the ranking results, the priority marking schemes are extracted from high to low to generate preliminary candidate decision combinations; By combining the generation rules, the preliminary candidate decision combinations are constrained and optimized to generate the optimized combination; Verify the stability index and execution efficiency value of the optimized combination. When the comprehensive score difference meets the preset threshold range, the final candidate decision combination is determined based on the principle of giving priority to the solution with low resource utilization.

[0013] In a preferred embodiment, the step S6 specifically includes: Obtain historical performance data and evaluation data of the candidate decision combination under historically similar environments, and perform data cleaning and formatting to obtain a structured historical performance data set; The structured historical performance dataset combines environmental changes and core features, and uses a random forest model to perform feature importance analysis on historical performance data to identify the key influencing factors of the current environment; Based on key influencing factors and the data characteristics of the current environment, real-time environmental data streams are obtained to determine environmental change trends. If the change trend exceeds the preset threshold, the decision-making combination is prioritized. The adjusted decision-making combination is combined with the dimensions of environmental adaptation and effect comparison to obtain decision-making performance data in similar historical environments. After comparison, the applicability ranking result is determined; Based on the applicability ranking results, multi-dimensional evaluation indicators are obtained based on the characteristics of candidate solutions and data associations, and a weighted linear regression algorithm is used to calculate the applicability score of the candidate decision combination; Generate decision recommendations based on the applicability scores and determine the applicability evaluation results of the decision combination in the current environment.

[0014] The preferred solution further includes S7: calculating the decision maintenance time of each solution in the candidate decision combination according to the applicability evaluation result, and determining the final decision execution time; The S7 specifically includes: The applicability assessment results are analyzed to obtain the solution matching degree. Based on the comparison of the solution matching degree with the preset threshold line, a preliminary matching degree judgment result is obtained, and further judgment is made: If the preset threshold is exceeded, the duration calculation method is used to analyze the decision maintenance time and obtain a preliminary maintenance time value; if the maintenance time value is lower than the average level, it is extended to obtain adjusted maintenance time data; the adjusted maintenance time data is double-checked in combination with the solution analysis method to determine whether it meets the execution conditions and determine the final execution time value; Based on the execution time value and the characteristics of the solutions in the candidate decision group, if there are multiple conflicting solutions, a decision execution sequence is generated by priority sorting; Performing time sequence allocation on the decision execution sequence, and verifying the rationality of the allocation using a support vector machine model to determine the final execution time allocation plan; Obtain the final execution time allocation plan, store the execution time constant of each plan in the system log, and update the evaluation result set to complete the decision execution process.

[0015] In the preferred solution, the calculation formula of the duration calculation model in S7 is: Initial duration = basic duration × (scheme matching degree / 100); Extension duration = initial duration × 0.2; Among them, the basic duration is preset to 30 days, and the solution matching degree is a rating value of 0-100.

[0016] In a second aspect, an embodiment of the present invention provides a business process control system based on a decision engine, including: The data collection module is used to collect environmental change data in business processes in real time and classify it into significant change groups and minor fluctuation groups; The data comparison module is used to compare the data of the small fluctuation group with the preset threshold. If it is lower than the preset threshold, it will be saved to the environmental change core set; The feature extraction module is used to extract features based on the data features of the significant change group and the core set of environmental changes, and conduct strategy matching analysis in combination with the decision optimization algorithm to generate a set of decision adjustment solutions; The decision evaluation module is used to evaluate the decision adjustment plan set using a historical analysis model, compare it with historical decision effect data, and obtain the execution efficiency value and stability index of each plan; The candidate decision module is used to comprehensively sort the decision adjustment scheme set according to the execution efficiency value and stability index, and select the schemes with stability index higher than the preset benchmark value as candidate decision combinations; The applicability evaluation module is used to combine the historical decision effect data corresponding to the candidate decision combination with the core set characteristics of the environmental changes to obtain the applicability evaluation results of the candidate decision combination in the current environment.

[0017] The embodiment of the present invention provides a control method and system for business processes based on a decision engine. By collecting environmental change data in the business process in real time, the data is classified, processed and filtered, and divided into a significant change group and a small fluctuation group to obtain a core set of environmental changes; a strategy matching analysis is performed on the significant change group to determine a set of adapted decision adjustment schemes; then, the execution efficiency and stability of each scheme are evaluated using historical execution records, and a comprehensive ranking is performed to determine a candidate decision combination; finally, the applicability of the candidate scheme is evaluated in combination with environmental characteristics. The present invention achieves a rapid response to environmental changes, improves resource utilization, optimizes the decision-making process, improves the accuracy and execution efficiency of decisions, improves the applicability of complex and changing business environments, and achieves dynamic optimization and continuous improvement of decisions, providing enterprises with more flexible and efficient decision support. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 This is a flowchart of a method for controlling a business process based on a decision engine provided by an embodiment of the present invention; Figure 2 It is a structural diagram of a control system for a business process based on a decision engine provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0020] Example 1 Reference Figure 1 , an embodiment of the present invention provides a method for controlling a business process based on a decision engine, comprising the following steps: S1: Collect environmental change data in business processes in real time and classify them according to the size of the environmental changes, dividing them into significant change group and minor fluctuation group.

[0021] S2: Compare the data of the small fluctuation group with the preset threshold and filter it. If it is lower than the preset threshold, it is marked as a data group that does not need to be adjusted, and the filtered environmental change core set is obtained.

[0022] S3: After obtaining the data features of the significant change group from the core set of environmental changes, the significant change group is subjected to strategy matching analysis using a decision optimization algorithm to generate a decision adjustment solution set.

[0023] S4: Use the historical analysis model to evaluate the decision adjustment plan set, compare it with the historical decision effect data, and obtain the execution efficiency value and stability index of each plan.

[0024] S5: Based on the execution efficiency value and stability index, the decision adjustment scheme set is comprehensively ranked, and the schemes with stability index higher than the preset benchmark value are selected as candidate decision combinations.

[0025] S6: Combine the historical decision effect data corresponding to the candidate decision combination with the core set characteristics of the environmental change to obtain the applicability evaluation results of the candidate decision combination in the current environment.

[0026] S7: Calculate the decision maintenance time of each plan in the candidate decision combination based on the applicability evaluation results, and determine the final decision execution time.

[0027] This embodiment collects environmental change data from business processes in real time and performs classification processing and filtering, dividing it into significant change groups and small fluctuation groups to obtain a core set of environmental changes; achieves accurate perception of environmental changes; performs strategy matching analysis on the significant change group to determine a set of adaptive decision adjustment plans; then, uses historical execution records to evaluate the execution efficiency and stability of each plan, performs comprehensive sorting, and determines a candidate decision combination; finally, evaluates the applicability of the candidate plan in combination with environmental characteristics, and calculates the decision maintenance time. The present invention realizes dynamic optimization management of business processes through intelligent environmental perception, multi-dimensional decision evaluation and adaptive execution control, combines historical effect data with multi-dimensional evaluation models, improves the matching accuracy and execution efficiency of decision plans, optimizes the decision process, improves the applicability of complex and changing business environments, and realizes dynamic optimization and continuous improvement of decisions.

[0028] The following is a detailed description of steps S1-S7.

[0029] In the preferred solution, step S1 obtains the classified environmental change dataset to obtain basic environmental information for subsequent analysis, specifically including: A raw environmental dataset containing temperature, humidity, and vibration parameters is collected in real time through a sensor network. If the raw environmental dataset contains missing values, the missing values ​​are addressed using the mean-filling method to obtain a complete environmental dataset. The complete environmental dataset is classified using the K-means clustering algorithm, dividing the data into a significant change group and a small fluctuation group, resulting in a classified environmental change dataset. Based on the classified environmental change dataset, the feature vectors of the significant change group are extracted to generate a significant change feature set. Principal component analysis and dimensionality reduction are performed on the significant change feature set to obtain basic environmental information for subsequent analysis. If the dimensionality of the basic environmental information is below a preset threshold, a feature selection algorithm is used to filter highly correlated features to generate optimized basic environmental information. Time series analysis is used to extract trend analysis trends from the optimized basic environmental information, resulting in an environmental change trend dataset.

[0030] For example, temperature, humidity, and vibration sensors are deployed in a smart manufacturing workshop. Data is collected once per second to generate a raw data set containing temperature (unit: °C), humidity (%), and vibration acceleration (m / s²). Suppose that at a certain moment, the collected temperature is 25.3 °C, the humidity is 60.2%, and the vibration is 0.05 m / s². A preliminary classification process was performed on these data using the K-means clustering algorithm (K=2). Using Euclidean distance as the metric, the data were divided into a significant change group and a small fluctuation group. The thresholds for significant changes were set at ±2°C for temperature change, ±5% for humidity change, and ±0.1m / s² for vibration change. The distance between each data point and the initial cluster center was calculated. For example, the distance between a data point (26.5°C, 62.1%, 0.15m / s²) and the center of the significant change group (27.0°C, 63.0%, 0.20m / s²) was √((26.5-27.0)²+(62.1-63.0)²+(0.15-0.20)²)=0.95, which placed it in the significant change group.

[0031] After classification, the significant change group contains 10% of the data, such as records with temperature changes exceeding 27.3°C or vibrations exceeding 0.15 m / s². The minor fluctuation group contains 90% of the data, such as records with temperatures between 23.3°C and 27.3°C. The classified environmental change dataset is generated and a structured JSON file containing timestamps, classification labels, and raw data is generated, for example, {"timestamp":"2025-06-20 08:55:00", "label":"Significant change", "data":{"temp":26.5, "humidity":62.1, "vibration":0.15}} is stored in the database.

[0032] Based on this dataset, basic environmental information was extracted and statistical characteristics of the significant change group were calculated, such as the mean temperature of 26.8°C and standard deviation of 0.4, the mean humidity of 61.5% and standard deviation of 0.8, and the mean vibration of 0.18 m / s² and standard deviation of 0.03. This information was used for subsequent analysis, such as equipment failure prediction. Combined with the equipment operating status in the business process (such as a speed of 5000 rpm), if the proportion of data in the significant change group exceeded 15%, an alert was triggered and pushed to the monitoring system, improving production stability.

[0033] In the preferred solution, step S2 filters the small fluctuation group using a preset threshold value based on the classified environmental change data set. If the data value of the small fluctuation group is lower than the preset threshold value, it is marked as a data group that does not need to be adjusted, and the filtered environmental change core set is obtained, which specifically includes: Obtain the data groups classified in step S1, perform threshold judgment on the data groups with small fluctuations, and if the data value in a certain data group is lower than the preset threshold, mark it as a group that does not need to be adjusted, and obtain the marked data set. Through the marked data set, adopt a filtering processing mechanism to eliminate the data groups marked as not needing to be adjusted, and obtain a streamlined core data set of environmental changes. According to the streamlined core data set of environmental changes, obtain the change monitoring characteristics therein. Use the support vector machine algorithm to analyze the change trend in the core data set and determine the key change points. Based on the key change points, further group processing is performed on the core data set of environmental changes: if the fluctuation amplitude of a certain change point exceeds the preset range, it is classified as a data group that needs to be monitored, and obtain a grouped monitoring set. Based on the grouped monitoring set, obtain the distribution characteristics of the data groups that need to be monitored. Use information entropy to calculate and judge the change correlation between each group, obtain the final monitoring priority ranking, and then perform real-time data collection and update processing on the high-priority data groups to obtain the latest dynamic data set of environmental changes.

[0034] This example uses an environmental change dataset containing 1,000 records of ambient temperature changes as an example. Each record includes a timestamp and a temperature value ranging from -10 to 40 degrees Celsius. Initial classification divides the data into a small fluctuation group, a significant fluctuation group, and an abnormal fluctuation group. The small fluctuation group contains 400 records, indicating a temperature change of less than 2 degrees Celsius. Next, a filtering process is performed on the small fluctuation group, using a preset threshold of 1.5 degrees Celsius. An algorithm is used to iterate over each record, calculating the absolute value of the temperature change. If the change is less than 1.5 degrees Celsius, it is marked as "no adjustment required."

[0035] Table 1 Comparison of small fluctuation groups and thresholds The results are shown in Table 1. The proportion of the minor fluctuation group dropped from 40% to 15%, and the proportion of the significant fluctuation group increased to 50%. This optimized resource allocation, improved efficiency and accuracy, and provided a precise basis for subsequent environmental control strategies.

[0036] In the preferred solution, step S3 obtains data features of the significant change group for the filtered core set of environmental changes, uses a pre-established decision optimization algorithm to perform strategy matching analysis on the significant change group, and determines an adaptive decision adjustment solution set, specifically including: Obtain preliminary structural information of the core data set, determine the scope and category of significant changes, extract data features in the change group based on the scope and category of significant changes, quantify key indicators, obtain a feature vector set, and input it into the pre-built decision optimization model for analysis. If the change amplitude of the feature vector exceeds the preset threshold, the strategy matching process is triggered to determine the adaptation strategy direction.

[0037] Then, based on the results of the strategy matching process, we obtain the decision adjustment content corresponding to the change group, use classification technology to group the adjustment content, and determine the preliminary decision adjustment plan set, where: If the adjusted content after grouping is consistent with the trend of environmental changes, the solution details are further optimized to obtain a fine-tuned solution set.

[0038] For the fine-tuned solution set, obtain verification data related to the adaptation solution, judge the feasibility of the solution set through comparative analysis, and form the final output result.

[0039] In this embodiment, a principal component analysis algorithm is used to extract the key features of the significant change group, which include temperature change and humidity change. A multi-objective optimization genetic algorithm (NSGA-II) is used to calculate the fitness score of each significant change group. A decision tree model is used to match the feature vector with a preset strategy library.

[0040] Furthermore, for the core set of environmental changes filtered in step S2, a data screening algorithm is used to process the environmental change data.

[0041] In this embodiment, a core set containing 100 environmental variable data points is used as an example for explanation: Using a time series-based anomaly detection algorithm (such as the Z-score method), the change amplitude of each data point is calculated, the threshold is set to 2.5 standard deviations, and 20 data points with significant change amplitudes exceeding the threshold are screened out. If a variable rises from 50.2 to 75.8 within 24 hours, the change rate reaches 51%, and it is marked as a significant change.

[0042] Next, the data features of the significant change group were obtained, and the principal component analysis (PCA) algorithm was used to extract key features. Assuming that the analysis results showed that temperature changes contributed 60% of the variance and humidity changes contributed 30%, these two features were used as the main analysis dimensions to generate a feature vector matrix.

[0043] Subsequently, a multi-objective genetic algorithm was used to optimize the model. The algorithm input the feature vectors and a historical decision database (containing 500 historical records). After 100 generations of iterative optimization, the fitness score for each significantly changed group was calculated to be 0.85, exceeding the average of 0.7 for all other groups. Further strategy matching analysis was performed on the significantly changed groups. Using a decision tree model, the feature vectors were matched to a pre-defined strategy library (containing 10 strategies, such as cooling and humidification). For example, the group dominated by temperature changes was matched to a "5-degree cooling" strategy, and the group dominated by humidity changes was matched to a "10% humidification" strategy. These strategy combinations were generated. Finally, a set of suitable decision adjustment options was determined. The matching strategies were comprehensively evaluated using a weighted scoring mechanism (weighting temperature at 0.6 and humidity at 0.4). The final set of three options, including "5-degree cooling + 5% humidification," received a comprehensive score of 0.9. Option 1, which included "5-degree cooling + 5% humidification," also output detailed execution parameters, such as 80% operating power for the cooling device and a 2-hour operating time for the humidification device, enabling the system to automatically execute the options.

[0044] The above process forms a complete logical chain through data analysis, feature extraction, optimization algorithm and strategy matching, and relies on information technology to achieve automated processing, thereby improving the efficiency and accuracy of responding to environmental changes.

[0045] In the preferred solution, step S4 specifically includes: Extract historical decision-making effect data from historical execution records, perform data cleaning, remove outliers and missing values, and obtain a standardized decision-making effect data set; Construct a feature vector including execution time, resource consumption and result quality to obtain the solution feature set; The random forest algorithm is used to classify the feature set of the schemes to obtain the execution efficiency value and stability index of each scheme; Among them, the efficiency classification result is obtained according to the execution efficiency value of the plan. If the efficiency classification result is lower than the preset threshold, the failure case data is extracted from the historical execution record, analyzed and adjusted and optimized, and then the historical analysis model is used to simulate the execution and calculate the stability index.

[0046] In this embodiment, when extracting decision effect data from historical execution records, relevant records can be obtained from the relational database through the database query language SQL, assuming that the database contains fields "decision ID", "execution time", "execution result" and "resource consumption".

[0047] For example, using the SQL statement "SELECT decision ID, execution result, resource consumption FROM history records WHERE execution time BETWEEN '2024-01-01' AND '2024-12-31'" to extract data for 2024 yields 1,000 records, each containing the decision success rate (e.g., 90%) and resource consumption (e.g., 500MB of memory). For a decision adjustment plan set with three plans (A, B, and C), a script automatically extracts plan parameters from the plan library. For example, the resource allocation ratio for plan A is 0.6, for plan B it is 0.5, and for plan C it is 0.7.

[0048] When using the historical analysis model to evaluate solution performance, the random forest algorithm is used. The input features include resource consumption, success rate, and execution time, and the output is the execution efficiency value and stability index: Taking Solution A as an example, the training data consists of 500 historical execution records, with a feature vector of [resource consumption: 500MB, success rate: 0.9, execution time: 10s]. A random forest model (100 trees, maximum depth 10) predicted an execution efficiency of 0.85 (ranging from 0 to 1, with 1 being the highest efficiency), and a stability index with a standard deviation of 0.05 (based on 10 repeated predictions). The analysis process included data preprocessing (normalizing resource consumption to [0, 1]), model training (80% training set, 20% test set, achieving 95% accuracy), and result verification. This process was repeated for Solutions B and C, resulting in efficiency values ​​of 0.78 and 0.07 for Solution B, and 0.82 and 0.06 for Solution C.

[0049] Finally, comparing efficiency and stability, solution A was recommended due to its high efficiency and low standard deviation.

[0050] Step S4 can be automated using a Python script, calling pandas to process data and scikit-learn to train the model.

[0051] In the preferred embodiment, step S5 specifically includes: If the stability index is higher than the preset threshold, the corresponding solution will be marked as a priority.

[0052] The marked decision adjustment plan set is comprehensively ranked using a weighted scoring algorithm that combines stability indicators with execution efficiency values.

[0053] According to the ranking results, priority marking schemes are extracted from high to low to generate preliminary candidate decision combinations.

[0054] By combining generation rules, the preliminary candidate decision combinations are constrained and optimized to generate the optimized combination.

[0055] Verify the stability index and execution efficiency value of the optimized combination. When the comprehensive score difference meets the preset threshold range, the final candidate decision combination is determined based on the principle of giving priority to the solution with low resource utilization.

[0056] In this embodiment, it is assumed that three decision adjustment schemes A, B, and C are obtained through step S4, and their execution efficiency values ​​are 0.85, 0.78, and 0.92 respectively, and their stability indicators are 0.90, 0.95, and 0.88 respectively, and the preset stability benchmark value is 0.89.

[0057] First, a comprehensive ranking is performed based on the execution efficiency and stability indicators. A weighted scoring algorithm is used, with the weight distribution being 60% for execution efficiency and 40% for stability. The comprehensive score for execution efficiency is calculated as follows: ; The ranking result is C>A>B. Next, we screened the solutions with stability index higher than 0.89. Solutions A and B met the criteria and were marked as priority solutions, while C (0.88) did not meet the criteria.

[0058] We further analyzed the priority options A and B, combined with business scenarios (such as system resource utilization, assuming A is 20% and B is 30%), and selected the option with lower resource utilization to optimize system performance. We ultimately determined A as the candidate decision combination.

[0059] In this embodiment, step S5 can be implemented through a Python script, which inputs efficiency and stability data, automatically calculates weighted scores, sorts and screens priority solutions, and outputs the final decision. The decision result can be directly applied to system optimization.

[0060] In the preferred embodiment, step S6 specifically includes: For each decision combination, the corresponding historical effect data and evaluation data are obtained from the pre-established database, and the data is cleaned and formatted to obtain a structured historical performance data set.

[0061] The structured historical performance dataset combines environmental changes and core features, and uses a random forest model to perform feature importance analysis on historical performance data to determine the key influencing factors of the current environment.

[0062] The key influencing factors are combined with the data characteristics of the current environment to obtain real-time environmental data streams and judge the trend of environmental changes. If the trend exceeds the preset threshold, the priority of the decision combination is adjusted.

[0063] The adjusted decision combination is combined with the dimensions of environmental adaptation and effect comparison to obtain decision performance data in similar historical environments. After comparison, the applicability ranking results are determined.

[0064] According to the applicability ranking results, multi-dimensional evaluation indicators are obtained based on the characteristics of candidate solutions and data associations, and a weighted linear regression algorithm is used to calculate the applicability score of the candidate decision combination.

[0065] Generate decision recommendations based on the applicability scores and determine the applicability evaluation results of the decision combination in the current environment.

[0066] Furthermore, in combination with the requirements of the evaluation conclusion, if the comprehensive score of the applicability score is lower than the preset threshold, the alternative plan screening process is triggered to obtain the data of the alternative decision combination, judge its matching degree in the current environment, and determine the optimal alternative plan.

[0067] In this embodiment, the historical effect evaluation data of the final candidate decision combination is obtained and its applicability in the current environment is judged by combining the core set characteristics of the environmental change. The specific implementation method is as follows: First, let's assume a candidate decision combination A, consisting of Strategy 1 (60% advertising placement) and Strategy 2 (15% promotional discount). The system database retrieves historical performance data from the past 12 months under similar market conditions, revealing an average sales growth rate of 8.5% and an increase in customer satisfaction ratings to 4.2 out of 5.0. This data was acquired using an SQL query algorithm, with filtering criteria based on time range, similar market size (±10% error), and consumer demographic characteristics (±5% error in age distribution) to ensure data accuracy.

[0068] Next, we analyzed the core set of characteristics of the current environmental changes, including a 5% decline in the economic index, a 3 percentage point drop in the consumer confidence index, and a 20% increase in competitive product activity. We constructed a model using feature vectors, comparing historical data with the current environmental characteristics. We used a weighted linear regression algorithm to calculate the suitability score, with weights of 0.4 for the economic index, 0.3 for consumer confidence, and 0.3 for competitive product activity. This yielded a suitability score of 0.72 (out of a maximum score of 1.0) for Portfolio A. Further analysis of the score composition revealed that the decline in the economic index had a significant impact on Portfolio A, resulting in a 0.2 point decrease in its score. However, the promotional discount strategy partially offset this negative impact, increasing the score by 0.05.

[0069] Finally, based on the applicability assessment results, the system automatically generates a report, indicating that combination A has medium applicability in the current environment. It recommends optimizing the advertising ratio to 50% to reduce cost risks. Combined with real-time competitor monitoring data (assuming the competitor discount rate increases to 20%), the discount is dynamically adjusted to 18%. The adjustment plan is automatically pushed to the execution module through the decision support system, forming a closed-loop optimization logic, which improves decision-making adaptability and aligns with business goals.

[0070] In the preferred embodiment, step S7 specifically includes: Obtain the data of each solution in the candidate decision group, perform a preliminary analysis on the suitability evaluation result set, analyze the suitability evaluation result to obtain the solution matching degree, and obtain a preliminary matching degree judgment result by comparing the solution matching degree with the preset threshold line, and further judge: If it exceeds the preset threshold line, the duration calculation method is used to analyze the decision maintenance duration and obtain a preliminary maintenance duration value; if the maintenance duration value is lower than the average level, it is extended to obtain the adjusted maintenance duration data; combined with the solution analysis method, the adjusted maintenance duration data is double-checked to determine whether it meets the execution conditions and determine the final execution duration value.

[0071] According to the execution time value and the characteristics of the solutions in the candidate decision group, if there are multiple solutions conflicting, the decision execution sequence is generated by priority sorting.

[0072] The decision execution sequence is time-series allocated, and the rationality of the allocation is verified using a support vector machine model to determine the final execution time allocation plan.

[0073] Obtain the final execution time allocation plan, store and record the execution time constant of each plan, store it in the system log and update the evaluation result set to complete the decision execution process.

[0074] In the preferred solution, the calculation formula of the duration calculation model in step S7 is: Initial duration = basic duration × (scheme matching degree / 100); Extension duration = initial duration × 0.2; Among them, the basic duration is preset to 30 days, and the solution matching degree is a rating value of 0-100.

[0075] In this embodiment, the decision combination in step S6 is obtained, the decision maintenance time of each solution is calculated based on the applicability evaluation results, and the final execution time is adjusted based on the matching degree. First, the system receives candidate decision combination data, which includes three solutions, as shown in Table 2 below: Table 2 Decision maintenance time of each plan calculated based on the applicability evaluation results In this example, the system allocates the time proportionally: the total of 30.6 + 23.4 + 33.12 = 87.12 days, which falls within the upper limit and is therefore output directly. Plan C, due to its high degree of compatibility, receives the longest execution time and is therefore prioritized. The system stores the result in JSON format: {A:30.6, B:23.4, C:33.12}, and pushes it to the decision execution module. This entire process is automated through algorithms, ensuring logical rigor and efficiency.

[0076] Example 2 Further explanation is given in conjunction with Example 1. Figure 2 , provides a business process control system based on a decision engine, including: The data collection module is used to collect environmental change data in business processes in real time and classify them into significant change groups and minor fluctuation groups.

[0077] The data comparison module is used to compare the data of the small fluctuation group with the preset threshold. If it is lower than the preset threshold, it is saved to the environmental change core set.

[0078] The feature extraction module is used to extract features based on the data features of the significant change group and the core set of environmental changes, and to perform strategy matching analysis in combination with the decision optimization algorithm to generate a set of decision adjustment plans.

[0079] The decision evaluation module is used to evaluate the decision adjustment plan set using a historical analysis model, compare it with historical decision effect data, and obtain the execution efficiency value and stability index of each plan.

[0080] The candidate decision module is used to comprehensively sort the decision adjustment scheme set according to the execution efficiency value and stability index, and select the scheme with stability index higher than the preset benchmark value as the candidate decision combination.

[0081] The applicability evaluation module is used to combine the historical decision effect data corresponding to the candidate decision combination with the core set characteristics of the environmental changes to obtain the applicability evaluation results of the candidate decision combination in the current environment.

[0082] The above embodiments are merely preferred technical solutions of the present invention and should not be construed as limiting the present invention. The scope of protection of the present invention shall be the technical solutions set forth in the claims, including equivalent alternatives to the technical features of the technical solutions set forth in the claims. In other words, equivalent alternatives and improvements within this scope are also within the scope of protection of the present invention.

Claims

1. A business process control method based on a decision engine, characterized in that: The following steps are involved: S1: Collects environmental change data in business processes in real time and classifies it into significant change group and minor fluctuation group according to the size of the environmental change; S2: Compare the data of the small fluctuation group with the preset threshold and filter it. If it is lower than the preset threshold, it is marked as a data group that does not need to be adjusted, and the filtered environmental change core set is obtained; S3: After obtaining the data features of the significant change group from the core set of environmental changes, the significant change group is then subjected to strategy matching analysis using a decision optimization algorithm to generate a set of decision adjustment solutions; S4: Use the historical analysis model to evaluate the decision adjustment plan set, compare it with the historical decision effect data, and obtain the execution efficiency value and stability index of each plan; S5: Based on the execution efficiency value and stability index, the decision adjustment plan set is comprehensively ranked, and the plan with a stability index higher than the preset benchmark value is selected as the candidate decision combination; S6: Combine the historical decision effect data corresponding to the candidate decision combination with the core set characteristics of the environmental changes to obtain the applicability evaluation results of the candidate decision combination in the current environment.

2. The method for controlling a business process based on a decision engine according to claim 1, characterized in that: Said S1 specifically includes: Collect raw environmental data sets containing temperature, humidity, and vibration parameters in real time through a sensor network; The K-means clustering algorithm was used for classification processing to divide the data into a significant change group and a small fluctuation group, and the classified environmental change data set was obtained; According to the classified environmental change data set, the feature vectors of the significant change group are extracted to generate the significant change feature set; Principal component analysis is performed on the significant change feature set, and after dimensionality reduction processing, it is compared with the preset threshold. If it is lower than the preset threshold, high-correlation features are screened through feature selection algorithm, and then time series analysis method is used to extract environmental change trends.

3. The method for controlling a business process based on a decision engine according to claim 1, characterized in that: Said S2 specifically includes: A threshold judgment is performed on the small fluctuation group. If the data value in a certain data group is lower than the preset threshold, it is marked as not requiring adjustment and eliminated to obtain the core set of environmental changes; The support vector machine algorithm is used to identify key change points in the core concentration of environmental changes, and change points that exceed the preset range are classified as key monitoring data groups; Obtain the distribution characteristics of key monitoring data groups, use information entropy calculation to determine the correlation between changes in each group, and obtain the final monitoring priority ranking; Real-time data collection and update processing are carried out for high-priority data groups to obtain the latest dynamic data sets of environmental changes.

4. The method for controlling a business process based on a decision engine according to claim 1, characterized in that: The S3 specifically includes: After obtaining preliminary structural information of the core data set and determining the scope and categories of significant changes, feature extraction is performed and key indicators are quantified to obtain a feature vector set. This is then input into a pre-built decision optimization model for analysis. If the magnitude of the feature vector change exceeds a preset threshold, the strategy matching process is triggered to determine the appropriate strategy direction. Then, based on the results of the strategy matching process, the decision adjustment content corresponding to the change group is obtained, and the adjustment content is grouped using classification technology to determine the decision adjustment plan set, where: If the adjustments after grouping are consistent with the trend of environmental changes, the plan will be further optimized; Obtain verification data related to the current adaptation plan, compare it with the optimized plan, judge the feasibility of the plan set, and obtain the final decision adjustment plan set.

5. The method for controlling a business process based on a decision engine according to claim 1, characterized in that: The S4 specifically includes: Extract historical decision-making effect data from historical execution records, perform data cleaning, remove outliers and missing values, and obtain a standardized decision-making effect data set; Construct a feature vector including execution time, resource consumption and result quality to obtain the solution feature set; The random forest algorithm is used to classify the feature set of the schemes to obtain the execution efficiency value and stability index of each scheme.

6. The method for controlling a business process based on a decision engine according to claim 1, characterized in that: The S5 specifically includes: If the stability index is higher than the preset threshold, the corresponding solution will be marked as a priority; The marked decision adjustment plan set is comprehensively ranked using the stability index combined with the execution efficiency value through a weighted scoring algorithm; According to the ranking results, the priority marking schemes are extracted from high to low to generate preliminary candidate decision combinations; By combining the generation rules, the preliminary candidate decision combinations are constrained and optimized to generate the optimized combination; Verify the stability index and execution efficiency value of the optimized combination. When the comprehensive score difference meets the preset threshold range, the final candidate decision combination is determined based on the principle of giving priority to the solution with low resource utilization.

7. The method for controlling a business process based on a decision engine according to claim 1, characterized in that: The S6 specifically includes: Obtain historical performance data and evaluation data of the candidate decision combination under historically similar environments, and perform data cleaning and formatting to obtain a structured historical performance data set; The structured historical performance dataset combines environmental changes and core features, and uses a random forest model to perform feature importance analysis on historical performance data to identify the key influencing factors of the current environment; Based on key influencing factors and the data characteristics of the current environment, real-time environmental data streams are obtained to determine environmental change trends. If the change trend exceeds the preset threshold, the decision-making combination is prioritized. The adjusted decision-making combination is combined with the dimensions of environmental adaptation and effect comparison to obtain decision-making performance data in similar historical environments. After comparison, the applicability ranking result is determined; Based on the applicability ranking results, multi-dimensional evaluation indicators are obtained based on the characteristics of candidate solutions and data associations, and a weighted linear regression algorithm is used to calculate the applicability score of the candidate decision combination; Generate decision recommendations based on the applicability scores and determine the applicability evaluation results of the decision combination in the current environment.

8. The method for controlling a business process based on a decision engine according to claim 1, characterized in that: Also included is S7: calculating the decision maintenance time of each plan in the candidate decision combination according to the applicability evaluation result, and determining the final decision execution time; The S7 specifically includes: The applicability assessment results are analyzed to obtain the solution matching degree. Based on the comparison of the solution matching degree with the preset threshold line, a preliminary matching degree judgment result is obtained, and further judgment is made: If the preset threshold is exceeded, the duration calculation method is used to analyze the decision maintenance time and obtain a preliminary maintenance time value; if the maintenance time value is lower than the average level, it is extended to obtain adjusted maintenance time data; the adjusted maintenance time data is double-checked in combination with the solution analysis method to determine whether it meets the execution conditions and determine the final execution time value; Based on the execution time value and the characteristics of the solutions in the candidate decision group, if there are multiple conflicting solutions, a decision execution sequence is generated by priority sorting; Performing time sequence allocation on the decision execution sequence, and verifying the rationality of the allocation using a support vector machine model to determine the final execution time allocation plan; Obtain the final execution time allocation plan, store the execution time constant of each plan in the system log, and update the evaluation result set to complete the decision execution process.

9. The method for controlling a business process based on a decision engine according to claim 8, characterized in that: The calculation formula of the duration calculation model in S7 is: Initial duration = basic duration × (scheme matching degree / 100); Extension duration = initial duration × 0.2; Among them, the basic duration is preset to 30 days, and the solution matching degree is a rating value of 0-100.

10. A business process control system based on a decision engine, characterized in that: include: The data collection module is used to collect environmental change data in business processes in real time and classify it into significant change groups and minor fluctuation groups; The data comparison module is used to compare the data of the small fluctuation group with the preset threshold. If it is lower than the preset threshold, it will be saved to the environmental change core set; The feature extraction module is used to extract features based on the data features of the significant change group and the core set of environmental changes, and conduct strategy matching analysis in combination with the decision optimization algorithm to generate a set of decision adjustment solutions; The decision evaluation module is used to evaluate the decision adjustment plan set using a historical analysis model, compare it with historical decision effect data, and obtain the execution efficiency value and stability index of each plan; The candidate decision module is used to comprehensively sort the decision adjustment scheme set according to the execution efficiency value and stability index, and select the schemes with stability index higher than the preset benchmark value as candidate decision combinations; The applicability evaluation module is used to combine the historical decision effect data corresponding to the candidate decision combination with the core set characteristics of the environmental changes to obtain the applicability evaluation results of the candidate decision combination in the current environment.

Images