Intelligent order dispatching method based on multi-dimensional project accounting

Abstract

The invention relates to an intelligent order dispatching method based on multi-dimensional project accounting, and belongs to the technical field of computers.The method comprises the steps that multi-dimensional data are collected and preprocessed, and the multi-dimensional data comprise task dimension data and personnel dimension data; configuring the weight of each dimension in the task dimension data and the weight of each dimension in the personnel dimension data; dynamically adjusting the weight of each dimension according to a real-time scene; according to the weight of each dimension, weighting calculation is adopted to generate a matching degree score of the task of each person and the person; sorting from high to low according to the matching degree scores, selecting candidate persons, and screening out final order sending persons; and displaying the order sending result, verifying the order sending result, recording the actual completion result of each order sending, and generating a verification report. According to the method, the optimal order receiving can be automatically matched through the multi-dimensional accounting model, the management cost is reduced, and the task circulation efficiency is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of computers, and particularly relates to an intelligent task assigning method based on multi-dimensional project accounting. BACKGROUND

[0002] The current enterprise internal task assigning has the following problems:

[0003] 1. Multi-reliance on artificial experience (for example, project manager subjective assignment), and there are problems such as response lag, low efficiency, and resource mismatch.

[0004] 2. Only according to the current workload or experience of the staff, the key factors such as task attribute (urgency / complexity), skill matching degree of the staff, and regional distribution are ignored.

[0005] 3. It is impossible to dynamically adjust according to the real-time progress of the project, newly acquired skills of the staff, or sudden high-priority tasks.

[0006] 4. Artificial statistics of staff load and task time and effort are lack of quantitative data support, and it is easy to lead to uneven busy or overload operation. SUMMARY

[0007] In view of the above problems of the prior art, the purpose of the present application is to provide an intelligent task assigning method based on multi-dimensional project accounting, which automatically matches the optimal task assignment through a multi-dimensional accounting model, reduces the management cost, and improves the task flow efficiency.

[0008] The application provides an intelligent task assigning method based on multi-dimensional project accounting, which comprises the following steps:

[0009] S1, multi-dimensional data is collected and preprocessed, and the multi-dimensional data comprises task dimension data and personnel dimension data, wherein the task dimension data comprises task type label, task complexity, time cost and task priority, and the personnel dimension data comprises personnel skill label, historical performance quantitative value of the personnel, load coefficient and cooperation preference of the personnel;

[0010] S2, the weight i of each dimension in the task dimension data and the weight j of each dimension in the personnel dimension data are configured; wherein i=1~n, j=1~m; n is the number of task dimensions; m is the number of personnel dimensions; the weight of each dimension is dynamically adjusted according to the real-time scene, and the weight before adjustment and the weight after adjustment both satisfy the following conditions: The sum of the task dimension weights is 100%, and the sum of the personnel dimension weights is 100%;

[0011] S3, according to the weight of each dimension, a weighted calculation is performed to generate a matching degree score P of the task and the personnel of each personnel,

[0012]

[0013] wherein, the task dimension i is the quantified value of the i th task dimension; the weight i is the weight value of the i th task dimension; the personnel dimension j is the quantified value of the j th personnel dimension; the weight j is the weight value of the j th personnel dimension;

[0014] S4, after calculating the matching degree score of all personnel, ranking from high to low according to the matching degree score, selecting the top W personnel as the candidate personnel, and screening the final dispatch personnel according to whether the matching degree score is higher than the threshold value, whether the candidate personnel is abnormal and the multi-role coordination adaptation principle; wherein, W is the number of candidate personnel;

[0015] S5, displaying the dispatch result of the dispatch personnel, the dispatch result including: historical data of the task and the personnel and weight calculation process; verifying the dispatch result, recording the actual completion result of each dispatch, and generating a verification report.

[0016] Further, in S1, the pre-processing of multi-dimensional data includes:

[0017] S11, pre-processing of task dimension data, including: obtaining a task description document, parsing the content of the task description document to generate a task type label, predicting the task complexity, then calculating the time cost according to the task complexity, setting the priority correction coefficient of the task according to the preset priority rule and synchronizing the task priority;

[0018] S12, pre-processing of personnel dimension data, including: obtaining a personnel description document, parsing the content of the personnel description document to generate a personnel skill label, obtaining the historical performance quantified value, load coefficient and cooperation preference of the personnel.

[0019] Further, in S11, the task description text is extracted by natural language processing (NLP) technology, and the extracted keywords are matched according to the preset label library to generate corresponding task type labels;

[0020] A regression algorithm is used to construct a task complexity prediction model, wherein the input parameters of the task complexity model include the number of code lines, the number of function modules and the number of dependency relationships, and the output parameter of the task complexity model is the task complexity score;

[0021] The time cost is calculated according to the task complexity score, the personnel historical completion time length benchmark value and the priority correction coefficient.

[0022] Further, in S12, the personnel skill label is generated according to the personnel resume, training record and historical task skill usage statistics in the personnel description document;

[0023] Obtaining historical performance quantitative values of the personnel, including: calculating historical completion rate of the personnel on the task, generating task quality score and collaboration satisfaction;

[0024] According to the remaining working hours of the assigned task of the personnel collected in real time, the load coefficient is generated by comparing the accumulated value with the upper limit of the load of the personnel;

[0025] Recording the collaboration task data of the personnel and each collaboration role, calculating the collaboration success rate, and obtaining the collaboration preference of the personnel according to the ranking of the collaboration success rate with each collaboration role.

[0026] Further, in S2, the weights of the personnel dimension data of the personnel in different roles are configured differently according to the role difference of the personnel.

[0027] Further, in S2, the weights of each dimension are dynamically adjusted according to the real-time scene, including:

[0028] (1) Based on the rule engine, the weights of the corresponding dimensions are dynamically adjusted according to the preset adjustment rules, and the preset adjustment rules include: a preset trigger condition and an adjustment rule of the weight of the task dimension and the weight of the personnel dimension corresponding to the trigger condition, wherein the trigger condition includes: a change in task state, a change in personnel state, or a change in both;

[0029] (2) Based on feedback learning, the weights of the task dimension and the personnel dimension are automatically optimized through the correlation analysis of the historical task allocation results and the actual completion results.

[0030] Further, in S4, the candidate personnel are screened, including:

[0031] (1) Automatically excluding the candidate personnel with abnormal load coefficient;

[0032] (2) Judging the matching degree score of the candidate personnel, and triggering manual intervention when the matching degree score is lower than the preset matching degree threshold;

[0033] (3) According to the role of the personnel initiating the task allocation, the personnel who are task-collaboration-adapted with the personnel initiating the task allocation are preferentially selected.

[0034] Further, in S4, the scores of each dimension of the task dimension and the personnel dimension are displayed, and a matching degree ranking table is displayed, the matching degree ranking table including: the comprehensive score of each candidate personnel and the contribution value of each dimension.

[0035] Further, when it is detected that the task dimension data or the personnel dimension data is missing, a completion operation of the missing dimension data is automatically triggered;

[0036] When it is detected that the matching degree scores of all candidate personnel are lower than the preset matching degree threshold, a warning is issued, and an optimization direction suggestion for the current dispatching task is automatically pushed.

[0037] When it is detected that the personnel appear load overload, a warning is issued, and a list of alternative personnel is provided.

[0038] Further, when a network interruption is detected, data breakpoint continuation is supported; when it is detected that there is a dimension of data anomaly, the dimension is automatically excluded.

[0039] The present application has the following advantages:

[0040] (1) Task matching efficiency and accuracy are improved

[0041] Multi-dimensional data fusion: Traditional dispatching only relies on 1-2 core indicators, while the present application integrates task type, complexity, priority, time cost, and personnel skill, historical performance, load, collaboration preference, and more than 8 dimensions of data to build a more comprehensive evaluation system (for example, the matching degree formula includes the weighted sum of task dimensions and personnel dimensions), which improves the matching accuracy of tasks and personnel by 20%-35% compared with traditional methods.

[0042] Dynamic weight optimization: The weight is not a fixed value, but is dynamically adjusted according to the project phase (for example, the development phase focuses on skills, and the testing phase focuses on collaboration), the personnel state (for example, the load weight is automatically reduced when overloaded), or the administrator's needs (for example, the priority of urgent tasks is improved), avoiding "outdated matching" caused by static rules, and further improving the allocation efficiency. For example, after the load overrun warning is triggered, the system automatically adjusts the weight and recommends alternative personnel, reducing the probability of task backlog.

[0043] (2) Balance of resource utilization and team fairness

[0044] Load balancing control: By calculating the current load coefficient of personnel in real time and dynamically adjusting the weight, the system can automatically identify and avoid long-term overload of high-skilled personnel (for example, the weight is increased by 10% when the load is overrun), while reducing the idle of low-skilled personnel caused by too little task, and improving the overall load balancing degree. For example, the overload rate of high-skilled personnel is reduced from 40% to 10%, and the idle rate of low-skilled personnel is reduced from 25% to 5%.

[0045] Historical performance feedback mechanism: The system records the historical task completion of personnel (for example, on-time completion rate, quality score), and optimizes the weight through feedback learning (for example, the time cost weight is automatically increased due to the neglect of time cost leading to overtime), realizing the reasonable allocation between "high load but high skill" personnel and "low load but low skill" personnel, and avoiding the unfair phenomenon of "more work but no reward" for the capable.

[0046] (3) Management flexibility and system verifiability enhancement

[0047] Multi-role collaborative adaptation: Support the configuration of assignment preferences for different roles such as development managers, project managers, etc. (e.g., development manager template focuses on skill matching, project manager template focuses on time cost), meet the diversified management needs of enterprises. For example, in the assignment initiated by the development manager, the skill matching degree weight accounts for 50%, while in the assignment initiated by the project manager, the time cost weight accounts for 40%.

[0048] Visual verification and traceability: The assignment result interface displays the dimension scores (e.g., task type matching degree, personnel load coefficient) and weight calculation process (e.g., the current task type weight is 30%, and the personnel skill weight is 40%), supporting management personnel to trace back the allocation basis (e.g., clicking on the assignment record can view the input parameters of the task complexity prediction model, and the specific data of personnel historical performance), improving system transparency and credibility.

[0049] Exception handling and robustness: Through breakpoint continuation, fault tolerance mechanism (e.g., automatically complete or prompt when data is missing), version rollback (e.g., restore the original configuration when weight adjustment effect is not good), ensure the stable operation of the system in complex scenarios. For example, Git data collection is automatically resumed after interruption, or task type labels are completed through NLP when missing, avoiding assignment failure. Through exception handling technology, the manual intervention rate can be reduced from 40% to 15% or even less. Through system robustness, 99.99% availability can be ensured, such as automatically triggering weight adjustment when load exceeds limit.

[0050] (4) Technical expansibility and feasibility

[0051] Modular design: Data collection, weight adjustment, matching algorithm, result verification, etc. Modules are independently implemented, supporting subsequent function extension (such as adding "personnel geographic location" dimension to adapt to remote collaboration scenarios). Modular design can reduce the cost of function iteration. For example, the weight module can be upgraded separately.

[0052] Seamless integration with existing systems: Through RESTful API and Jira / Trello (task priority synchronization), Git (complexity prediction), WeChat Enterprise (collaboration satisfaction collection), etc. Mainstream tools are connected, without additional development of underlying systems by enterprises, reducing the landing cost, without the need to reconstruct enterprise infrastructure, shortening the deployment cycle. BRIEF DESCRIPTION OF DRAWINGS

[0053] The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and together with the description serve to explain the principles of the application. In the drawings:

[0054] Figure 1 A flow chart of an intelligent dispatching method based on multi-dimensional project accounting according to an embodiment of the present application;

[0055] Figure 2 A schematic diagram of a multi-dimensional data acquisition and quantification model architecture according to an embodiment of the present application;

[0056] Figure 3 A schematic diagram of a dispatching main flow and exception handling mechanism according to an embodiment of the present application;

[0057] Figure 4 A schematic diagram of a weight dynamic adjustment rule base according to an embodiment of the present application;

[0058] Figure 5 A schematic diagram of a matching degree calculation example according to an embodiment of the present application. DETAILED DESCRIPTION

[0059] In order to make the personnel in the art better understand the technical solutions in the embodiments of the present application, the technical solutions of the present application will be described clearly and completely below with reference to the drawings. Obviously, the described embodiments are some embodiments of the present application, rather than all the embodiments of the present application. It should be understood that these descriptions are only exemplary, and are not used to limit the scope of the present application. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor should belong to the scope of protection of the present application.

[0060] In addition, in the following description, the description of well-known structures and techniques has been omitted to avoid unnecessary confusion of the concepts disclosed in the present application.

[0061] In the description of the present application, it should be noted that, unless otherwise explicitly specified and limited, the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer" and the like indicate the orientation or positional relationship shown in the drawings, which are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application. In addition, the terms "first", "second", "third" are only for descriptive purposes and cannot be understood as indicating or implying relative importance. The terms "mounting", "connection", "connection" should be broadly understood, for example, it can be fixed connection, or detachable connection, or integral connection; it can be mechanical connection, or electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, or the communication between two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0062] The exemplary embodiments will be described in detail hereinbelow, examples of which are shown in the accompanying drawings. In the following description, the same numbers refer to the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of methods and systems consistent with some aspects of the present application as detailed in the appended claims.

[0063] The present application proposes an intelligent task assignment method based on multi-dimensional project accounting, which is an intelligent, adaptive and multi-target balanced task assignment method. The intelligent task assignment method of the present application can realize the following by combining the multi-dimensional accounting model and the dynamic weight adjustment mechanism:

[0064] 1. Through the multi-dimensional accounting model, the optimal single person is automatically matched. At least three dimensions (for example, task type, personnel skill matching degree, current workload, etc.) are defined, and a quantitative index system (for example, task complexity score, personnel historical completion rate value) is constructed.

[0065] 2. Support dynamic weight adjustment, automatically switch task assignment strategy according to task attributes (for example, urgent bug repair). According to the real-time data of the project (such as personnel load change, task priority adjustment), the weight coefficients of each dimension are dynamically updated, rather than fixed values.

[0066] 3. Use intelligent task assignment algorithm framework: clear algorithm input (multi-dimensional data), processing logic (weight calculation and matching rule), output (specific task assignment result). Establish a closed-loop learning mechanism of task assignment result→execution feedback, continuously optimize the accuracy of assignment; reduce management cost by more than 20%, and improve task flow efficiency by 30%.

[0067] As shown in Figure 1 and Figure 2 The present application proposes an intelligent task allocation method based on multi-dimensional project accounting, comprising the following steps:

[0068] S1, collecting and preprocessing multi-dimensional data, the multi-dimensional data comprising: task dimension data and personnel dimension data.

[0069] The task dimension data comprises task type label, task complexity, time cost and task priority. The personnel dimension data comprises personnel skill label, personnel historical performance quantitative value, load coefficient and personnel collaboration preference.

[0070] It should be noted that the present application integrates various task attributes in the task dimension data and various personnel attributes in the personnel dimension data. The task dimension data and the personnel dimension data are not limited to the above examples, and other attributes can be added, deleted, modified, etc. as needed.

[0071] In addition, the number of dimensions in the multi-dimensional data is at least 3, i.e. the number of dimensions is greater than or equal to 3.

[0072] The number of dimensions (n) contained in the task dimension data and the number of dimensions (m) contained in the personnel dimension data are both greater than or equal to 3, or n+m> 3.

[0073] In this step, the preprocessing of the multi-dimensional data comprises:

[0074] S11, preprocessing of the task dimension data, comprising: obtaining a task description document, parsing the content of the task description document to generate a task type label, predicting the task complexity, then calculating the time cost according to the task complexity, setting the priority correction coefficient of the task according to the preset priority rule and synchronizing the task priority.

[0075] (1) Task type label

[0076] The keywords are extracted from the task description text by natural language processing (NLP) technology, and the extracted keywords are matched according to the preset label library to generate the corresponding task type label.

[0077] In the present application, the keywords are, for example, backend development and interface debugging; the label library is, for example, development-backend and testing-automation.

[0078] That is, the keywords are extracted from the task description text by natural language processing (NLP) technology, and the extracted keywords are matched according to the preset label library to generate the corresponding task type label development-backend and testing-automation.

[0079] In addition, the present application also supports user-defined tags. For example, a development manager can add a "document-requirement analysis" tag.

[0080] (2) Task complexity

[0081] A regression algorithm is used to construct a task complexity prediction model. The input parameters of the task complexity model include the number of code lines, the number of functional modules, and the number of dependency relationships, and the output parameter of the task complexity model is the task complexity score.

[0082] In this step, a linear regression or random forest regression algorithm can be used to construct a complexity prediction model.

[0083] Among them, the number of code lines is extracted from a version control system such as Git, the number of functional modules is generated by a task decomposition module, and the number of dependency relationships is counted through a task association graph. The output is a complexity score ranging from 0 to 100 points.

[0084] (3) Time cost

[0085] The time cost is calculated according to the task complexity score, the historical completion time length benchmark value of the personnel, and the priority correction coefficient.

[0086] Specifically, based on the task complexity score (0-100 points) and the historical completion time length data of the personnel, the estimated completion time, i.e. the time cost, is calculated as:

[0087] Time cost = complexity score x historical time length benchmark value x priority correction coefficient (1)

[0088] Among them, the historical time length benchmark value: statistics the actual average completion time length of similar tasks in the past three months of this personnel. For example, the "development-backend" task historical average time length of personnel X = 3 days.

[0089] Priority correction coefficient: dynamically determined by the following rules:

[0090] S-level urgent task: coefficient = 0.8 (compress 20% estimated time length);

[0091] A-level important task: coefficient = 1.0 (keep benchmark time length);

[0092] B-level ordinary task: coefficient = 1.2 (increase 20% buffer time length).

[0093] (4) Task priority

[0094] The following two ways are used to realize task priority synchronization:

[0095] Automation synchronization: real-time docking with project management systems through API interfaces, automatically obtaining task priority labels (S / A / B levels). Among them, the project management system is, for example, a Jira or Trello system.

[0096] Manual intervention mechanism: support for development managers to manually override priority values (for example, set "priority = 90 points") and automatically convert to correction factors as follows:

[0097] Priority value ≥ 90 → correction factor = 0.8 (S level equivalent);

[0098] 80 ≤ priority value < 90 → correction factor = 1.0 (A level equivalent);

[0099] Priority value < 80 → correction factor = 1.2 (B level equivalent).

[0100] S12, preprocessing of personnel dimension data, including: obtaining personnel description documents, parsing the content of personnel description documents to generate personnel skill labels, obtaining historical performance quantization values, load coefficients and collaboration preferences of personnel.

[0101] (1) Personnel skill labels

[0102] According to the personnel resume, training records and historical task skill usage statistics in the personnel description document, the personnel skill labels are generated.

[0103] Among them, the detailed content of the personnel resume is extracted by OCR recognition technology, and the training records are obtained through system interface synchronization.

[0104] Historical task skill usage statistics, for example, personnel X uses Java in the "development-back end" task with an occupancy rate of 80%, and uses Python with an occupancy rate of 20%, and then generates a multi-level skill label: Java (master), Python (skilled).

[0105] (2) Historical performance quantization values of personnel

[0106] Obtain the historical performance quantization values of personnel, including: calculate the historical completion rate of personnel to the task, generate task quality score and collaboration satisfaction.

[0107] Historical completion rate: the number of tasks closed on time in the past 6 months / total number of tasks accepted) x 100%; Judgment standard: over 24 hours is considered not completed.

[0108] Task quality score: generated by test results or review scores, supporting multiple weights. For example, the test result is the number of bugs, and the review score is "quality score = 90 points", and the weights are respectively Bug number occupancy 60%, review score occupancy 40%.

[0109] Collaboration Satisfaction: Based on team member mutual evaluation data, calculate the average score.

[0110] Wherein, team member mutual evaluation data can use questionnaire star or enterprise internal collaboration platform score record.

[0111] The average score of collaboration satisfaction, for example, the collaboration satisfaction of personnel X is 85 points.

[0112] (3) Load coefficient

[0113] According to the real-time collection of personnel assigned task remaining working hours, after accumulation and comparison with the upper limit of the load of the personnel, the load coefficient is generated.

[0114] Load coefficient=(load upper limit-current load) / load upper limit (2)

[0115] The load coefficient is the ratio of the current load of the personnel to the upper limit, the value range is 0-1.0, which is used for quantitative scoring and early warning.

[0116] When the load coefficient is 1.0, it means that the personnel is in idle state. When the load coefficient>0.8, it triggers adjustment.

[0117] For example, real-time collection of personnel assigned task A remaining 2 hours, task B remaining 3 hours, after accumulation and comparison with the upper limit of the load of 8 hours per day, the load coefficient is generated as:

[0118] Load coefficient=(8-(2+3)) / 8=0.4

[0119] (4) Collaboration success rate

[0120] Record the collaboration task data of personnel and each collaboration role, calculate the collaboration success rate, and obtain the collaboration preference of the personnel according to the sorting of the collaboration success rate with each collaboration role.

[0121] Specifically, the collaboration success rate is the on-time completion rate of personnel and specific role collaboration, and the calculation formula is as follows:

[0122] On-time completion number / collaboration task number×100% (3)

[0123] Collaboration success rate can affect the weight of collaboration preference.

[0124] Collaboration preference modeling: record the collaboration task data of personnel and specific role, and calculate the collaboration success rate.

[0125] For example: record the collaboration task number of personnel and tester Y as 10, and the on-time completion number as 9, then the collaboration success rate is:

[0126] Collaboration success rate=9 / 10×100%=90% (4)

[0127] The application supports customizing preference labels, such as: "highly efficient in collaboration with tester Y".

[0128] S2, configure the weight i of each dimension in the task dimension data and the weight j of each dimension in the personnel dimension data, and dynamically adjust the weight of each dimension according to the real-time scene.

[0129] In the application, i = 1 ~ n, j = 1 ~ m; n is the number of task dimensions; m is the number of personnel dimensions. That is, the weight of each task attribute in the task dimension and the weight of each personnel attribute in the personnel dimension are configured.

[0130] In the application, a visual configuration interface (for example, a Web or mobile interface) is provided, and an administrator can adjust the weight of each dimension through a sliding bar or an input box. For example, the administrator sets the weight of the task type to 30% and the weight of the personnel skill to 40% through the visual configuration interface.

[0131] In this step, a role differentiation weight template is supported, and the weight of the personnel dimension data of personnel with different roles is differentiated according to the role difference of the personnel.

[0132] For example, the development manager template: skill matching degree accounts for 50%, historical performance accounts for 30%, and load accounts for 20%. The project manager template: time cost accounts for 40%, load accounts for 30%, and skill matching degree accounts for 30%.

[0133] According to the real-time scene, the weight of each dimension is dynamically adjusted, and the dynamic weight adjustment mechanism is as follows:

[0134] It should be noted that the weight before and after adjustment needs to meet the following conditions: That is, the sum of the weights of the task dimensions is 100%, and the sum of the weights of the personnel dimensions is 100%.

[0135] Specifically, according to the adaptive rule library for automatically updating the weight coefficients of each dimension according to the real-time scene, it is ensured that the sum of the weights of the task dimensions is 100% and the sum of the weights of the personnel dimensions is 100%.

[0136] That is, the weight adjustment limit: the weight value needs to be in the range of 0-100%, and the total sum is 100%. After the administrator adjusts, the system automatically normalizes the weight of other dimensions.

[0137] According to the real-time scene, the weight of each dimension is dynamically adjusted, including:

[0138] (1) Based on the rule engine, the weight of the corresponding dimension is dynamically adjusted according to the preset adjustment rule. The preset adjustment rule includes: a preset trigger condition and an adjustment rule of the weight of the task dimension and the weight of the personnel dimension corresponding to the trigger condition. Wherein, the trigger condition includes: task state change, personnel state change or both changes. For example: project phase change, personnel state anomaly, etc.

[0139] The specific rules of the weight adjustment strategy are illustrated as follows:

[0140] 1. Trigger condition = load exceeds limit for 3 consecutive days, weight adjustment strategy = (load weight + 10%, time cost weight - 5%)

[0141] That is, if the load of a certain personnel exceeds the limit for 3 consecutive days (for example, load coefficient ≤ 0.1), the load weight is automatically increased by 10% and the time cost weight is decreased by 5%.

[0142] 2. Trigger condition = task type proportion in current project is too high, weight adjustment strategy = task type weight - 5%

[0143] That is, if the proportion of a certain task type in the current project is too high (for example, the proportion of "development-backend" task is ≥ 70%), the task type weight is automatically decreased by 5% to balance the allocation demand of other types of tasks.

[0144] As shown in Figure 4 , specific examples of trigger conditions are as follows:

[0145] 1. Project phase change: dynamically adjust the task type weight. For example, from the development phase to the testing phase, increase the "test-automation" task weight by 10% and decrease the "development-backend" task weight by 5%.

[0146] 2. Personnel state anomaly: automatically decrease the historical performance weight and increase the time cost weight. For example, a certain personnel has been overtime for 2 consecutive days, the historical performance weight is decreased from 30% to 20%, and the time cost weight is increased from 10% to 20%.

[0147] 3. Administrator manual adjustment: input the adjustment value through the interface, and the system automatically normalizes the weight of other dimensions.

[0148] For example, the administrator temporarily increases the weight of "urgent task priority", inputs "urgent task priority weight from 20% to 30%" through the interface, and the system automatically normalizes the weight of other dimensions, with the total sum remaining 100%

[0149] (2) Based on feedback learning, the weight of the task dimension and the weight of the personnel dimension are automatically optimized through the correlation analysis of historical task allocation results and actual completion results.

[0150] For example, the actual effect this time is timeout due to neglecting time cost, then automatically optimize the weight, and increase the time cost weight by 5% next time.

[0151] It should be noted that the present application supports dynamic weight adjustment version rollback.

[0152] S3, according to the weight of each dimension, a weighted calculation is performed to generate a matching degree score P of each task and personnel,

[0153]

[0154] Wherein, n is the number of task dimensions (such as type, complexity, priority), m is the number of personnel dimensions (such as skill, historical performance, load). Task dimension i is the quantitative value of the i th task dimension; weight i is the weight value of the i th task dimension; personnel dimension j is the quantitative value of the j th personnel dimension; weight j is the weight value of the j th personnel dimension.

[0155] The present application uses quantitative score calculation for each dimension, and converts abstract attributes (such as task type matching degree, skill matching degree, load coefficient, etc.) into numerical scores.

[0156] For example, the task type matching degree is 0-100%, and the load coefficient is 0-1.0.

[0157] The following examples are illustrated in combination with specific indexes:

[0158] Task type matching degree: if the task type is "development-back end", and the personnel skill label contains "Java (expert)", then the matching degree is 80%. The proficiency level of the personnel skill label is quantified, such as "expert" = 1.0, "skilled" = 0.8, and "basic" = 0.5.

[0159] Historical performance coefficient: if the personnel historical on-time completion rate is 95%, then the coefficient is 0.95; if the task quality score is lower than 80 points, then the coefficient is reduced to 0.85.

[0160] Load coefficient: if the current load of the personnel is 7 hours (upper limit 8 hours), then the coefficient is 0.9. The lower the load, the higher the coefficient, and the load coefficient is inversely proportional to the current load. For example, when the current load of the personnel is 0, the load coefficient is 1.0 (the maximum value).

[0161] Reference Figure 5 The calculation of the matching degree score P is divided into two parts: task dimension and personnel dimension.

[0162] Task dimension score = (task type matching degree 30 points x 0.4) + (complexity score 70 points x 0.3) + (priority 90 points x 0.3) = 12 + 21 + 27 = 72 points;

[0163] Personnel dimension score = (personnel skill matching degree 80 points x 0.5) + (historical performance coefficient 0.95 x 0.3) + (load coefficient 0.125 x 0.2) = 40 + 0.285 + 0.025 = 0.31 (normalized and converted to percentage system, about 88 points);

[0164] Final matching degree score P = (task dimension score + personnel dimension score) / 2 = (72 + 88) / 2 = 80 points.

[0165] S4, after calculating the matching degree scores of all personnel, ranking them from high to low according to the matching degree scores, selecting the top W personnel as candidate personnel, and screening the final dispatch personnel according to whether the matching degree score is higher than the threshold, whether the candidate personnel is abnormal, and the multi-role coordination adaptation principle. Wherein, W is the number of candidate personnel. Preferably, W = 3. That is, ranking from high to low according to the matching degree score, selecting the top 3 personnel as candidate personnel.

[0166] It should be noted that in actual application, the number of candidate personnel W can support any number of candidate personnel, and the specific number is determined by user configuration.

[0167] This example selects 2 candidate personnel (A and B) for matching degree calculation demonstration, only for explaining the calculation logic. The actual system can support any number of candidate personnel, and the specific number is determined by user configuration.

[0168] Specifically, after selecting the dispatch candidate personnel, further screening the candidate personnel, including:

[0169] (1) Automatically exclude the candidate personnel with abnormal load coefficient.

[0170] Set the load coefficient threshold to 0.1, and when there is a load exceeding the threshold in the candidate personnel (i.e., load coefficient ≤ 0.1), it is automatically excluded.

[0171] (2) Judge the matching degree score of the candidate personnel, and when the matching degree score is lower than the preset matching degree threshold, trigger manual intervention.

[0172] Reference Figure 3 , set the matching degree threshold = 60 points, if there is no load abnormality in the candidate personnel, but the matching degree is lower than or equal to 60 points, trigger an alarm and suggest manual intervention. Wherein, triggering an alarm is to send an alarm prompt, for example, prompting "need to supplement skill tags or adjust task complexity".

[0173] (3) According to the role of the task initiator, the personnel who are suitable for the task of the task initiator are selected preferentially.

[0174] The specific logic of the multi-role collaborative adaptation is described as follows:

[0175] (3.1) When the development manager initiates the task, the personnel whose skill tags are highly consistent with the task requirements are matched preferentially. For example, when the task requirement initiated by the development manager is "Java development", the candidate personnel whose skill tags are "Java (expert)" are matched preferentially.

[0176] (3.2) When the project manager initiates the task, the candidate personnel with low load (for example, the load coefficient is greater than or equal to 0.8) and short time cost (for example, the estimated completion time is less than or equal to 2 days) are matched preferentially.

[0177] (3.3) If the task requirements initiated by the two roles conflict, the "role weight priority" rule is used to solve the problem.

[0178] For example, the development manager requires high-skilled personnel, and the project manager requires low-load personnel. The "role weight priority" rule is used to solve the problem, and the weight of the development manager requirement is configured to be 60%, and the weight of the project manager requirement is configured to be 40%.

[0179] In addition, the application can display the scores of each dimension of the task dimension and the personnel dimension, and display a matching degree ranking table. The matching degree ranking table includes the comprehensive score of each candidate personnel and the contribution value of each dimension.

[0180] For example, the matching degree of the task type is 80 points, the load of the personnel is 70 points, and the time cost is 60 points.

[0181] The comprehensive score of the candidate personnel and the contribution of each dimension are displayed. For example, the total score of the personnel X is 90 points, of which the contribution of the task type matching degree is 30 points, the contribution of the load balancing is 20 points, and the contribution of the time cost is 15 points.

[0182] S5, the task assignment result of the task assignment personnel is displayed, the task assignment result includes the historical data of the task and the personnel and the weight calculation process; the task assignment result is verified, the actual completion result of each task assignment is recorded, and a verification report is generated.

[0183] The application provides visualization and result verification, and each dimension score and weight calculation process are displayed through an interface, and the model is optimized based on actual completion data (for example, time deviation rate).

[0184] Reverse tracing function: by clicking a task assignment result, the historical data of the task and the personnel (for example, Figure 5Weight calculation process. Wherein, the historical data is for example the number of "development-backend" tasks completed by person X in the past 3 months and the average completion time. The weight is for example the current task type weight is 30%, and the person skill weight is 40%.

[0185] After generating the dispatch result, verification is performed, and the specific indicators of the result verification mechanism are as follows:

[0186] (1) Record the actual completion of each dispatch. For example: whether it is overtime, cost deviation rate, etc.

[0187] (2) The system generates a verification report regularly. For example, compare the deviation rate of "predicted completion time" and "actual completion time" every week, if the deviation rate > 15%, prompt to optimize the model or adjust the weight.

[0188] (3) Support artificial feedback correction: management personnel can score the dispatch result (for example, matching degree 85 points), and the system will feed back the data into the model training, further optimizing the matching logic.

[0189] The intelligent dispatch method based on multi-dimensional project accounting of the application provides functions such as exception handling, system robustness design and modular design

[0190] 1. Exception handling

[0191] For the automatic response process of scenarios such as data missing, low matching degree threshold, etc., such as label completion (NLP driven), load warning, etc., the artificial intervention rate can be reduced from 40% to 15% or even less.

[0192] Task and personnel data missing: when detecting that the task dimension data or the personnel dimension data is missing, the completion operation of the missing dimension data is automatically triggered.

[0193] If the task type label is missing, the "label completion" process is automatically triggered. For example, calling the NLP model to extract keywords from the task description and matching the label library.

[0194] If the personnel skill label is missing, the administrator is prompted to supplement. For example, the prompt "person Y has not labeled Python skill, manual confirmation is required" is issued.

[0195] When it is detected that the matching degree score of all candidate personnel is lower than the preset matching degree threshold, an early warning is issued, and the optimization direction suggestion for the current dispatch task is automatically pushed.

[0196] Specifically, if the matching degree of all candidate personnel is lower than the preset threshold (for example, the matching degree score is less than or equal to 60), a warning information is generated, and a possible optimization direction is automatically recommended. For example, the warning information is: "no suitable personnel at present, it is suggested to supplement skill labels or adjust task requirements". For example, the optimization direction is: "the task complexity is too high, and the task can be divided into subtasks".

[0197] When it is detected that the personnel has load overload, a warning is issued, and a list of alternative personnel is provided.

[0198] For example, when the current load of a certain personnel is 7.5 hours, the upper limit of the load is 8 hours, and the current load is close to the upper limit of the load, a warning is pushed, prompting "personnel X load is about to exceed the limit, it is suggested to adjust task allocation", and a list of alternative personnel is provided. For example, personnel with a load coefficient greater than or equal to 0.8 are provided to replace.

[0199] 2. Robustness of the system

[0200] The ability to maintain core functions under network interruption, data anomaly and other disturbances is embodied in fault tolerance mechanisms (for example, degradation when the algorithm module crashes) and breakpoint resume design.

[0201] When network interruption is detected, data breakpoint resume is supported, for example, automatic recovery of collection after network interruption.

[0202] When it is detected that there is a dimension with data anomaly, the matching algorithm supports fault tolerance mechanism, and automatically excludes the dimension. For example, if a certain dimension has data anomaly, the influence of the dimension is automatically excluded.

[0203] 3. Modular design

[0204] The present application independently realizes components such as data collection, weight adjustment and matching algorithm, supports low-coupling extension (for example, adding a "geographical location" dimension). Through modular design, the cost of function iteration can be reduced, and the present system can be seamlessly integrated. For example, through RESTful API, Jira, Git, enterprise WeChat and other tools are connected, data automatic synchronization (such as task priority, code line number) is realized, enterprise infrastructure does not need to be reconstructed, and deployment period is shortened.

[0205] The process of the intelligent task allocation method based on multi-dimensional project accounting of the present application will be described below with specific examples.

[0206] The running process example is as follows:

[0207] (1) The development manager creates a task in the system, "development user login module, priority S level".

[0208] (2) Automatically collect task type (development-backend), complexity (code line number estimation 500 lines, module number = 2).

[0209] Skill of the staff (Person A: Java expert, Person B: Java proficient), historical performance (Person A 95% on-time completion rate, Person B 85%), current load (Person A load = 7 hours, Person B = 5 hours).

[0210] (3) Calculate the matching degree as follows:

[0211] Person A matching degree = (task type matching degree 30 points x 0.4) + (complexity score 70 points x 0.3) + (priority 90 points x 0.3) + (personnel skill matching degree 80 points x 0.5) + (historical performance coefficient 0.95 x 0.3) + (load coefficient 0.125 x 0.2) = 90 points.

[0212] Person B matching degree = (task type matching degree 25 points x 0.4) + (complexity score 60 points x 0.3) + (priority 90 points x 0.3) + (personnel skill matching degree 70 points x 0.5) + (historical performance coefficient 0.85 x 0.3) + (load coefficient 0.375 x 0.2) = 85 points.

[0213] (4) Select the person with the highest matching degree, Person A, to assign the task, and record the basis for the assignment. That is, record the scores of each dimension and the weights.

[0214] If the subsequent load of Person A exceeds the limit, for example, the load = 8.5 hours after adding a new task, then automatically adjust the weights, for example, increase the load weight to 30%, recalculate the matching degree and trigger a warning. The warning prompt is "Person A load exceeds the limit, suggest adjusting the allocation".

[0215] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the embodiments of the present application, and not to limit them. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent substitutions for part of the technical features; and these modifications or substitutions do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present application should be covered within the protection scope of the present application.

Claims

1. A multi-dimensional project accounting-based intelligent dispatching method, characterized in that, The method comprises the following steps: S1, collecting and preprocessing multi-dimensional data, wherein the multi-dimensional data comprises task dimensional data and personnel dimensional data, wherein the task dimensional data comprises a task type label, task complexity, time cost and task priority, and the personnel dimensional data comprises a personnel skill label, a historical performance quantitative value of the personnel, a load coefficient and a collaboration preference of the personnel; S2, configuring the weight i of each dimension in the task dimension data and the weight j of each dimension in the personnel dimension data; wherein, i=1~n, j=1~m; n is the number of task dimensions; m is the number of personnel dimensions; dynamically adjusting the weight of each dimension according to the real-time scene, wherein the weight before adjustment and the weight after adjustment both satisfy the following conditions: The sum of the task dimension weights is 100%, and the sum of the personnel dimension weights is 100%. S3, generating a matching degree score P of each personnel according to the weights of each dimension by weighted calculation, wherein task dimension i is the quantitative value of the i th task dimension; weight i is the weight value of the i th task dimension; personnel dimension j is the quantitative value of the j th personnel dimension; weight j is the weight value of the j th personnel dimension; S4, after calculating the matching degree scores of all personnel, ranking them from high to low, selecting the top W personnel as candidate personnel, and screening the final personnel according to whether the matching degree score is higher than a threshold value, whether the candidate personnel is abnormal and the multi-role collaborative adaptation principle; wherein W is the number of candidate personnel; S5, displaying the dispatching result of the dispatched personnel, wherein the dispatching result comprises historical data of the task and the personnel and a weight calculation process; verifying the dispatching result, recording the actual completion result of each dispatching, and generating a verification report. 2.The intelligent dispatching method based on multi-dimensional project accounting according to claim 1, wherein, In S1, the preprocessing of the multi-dimensional data comprises: S11, preprocessing of task dimensional data, comprising: obtaining a task description document, parsing the content of the task description document to generate a task type label, predicting task complexity, then calculating time cost according to the task complexity, setting a priority correction coefficient of the task according to a preset priority rule and synchronizing the task priority; S12, preprocessing of personnel dimensional data, comprising: obtaining a personnel description document, parsing the content of the personnel description document to generate a personnel skill label, obtaining a historical performance quantitative value of the personnel, a load coefficient and a collaboration preference. 3.The intelligent dispatching method based on multi-dimensional project accounting according to claim 2, characterized in that, In S11, keywords are extracted from the task description text by natural language processing (NLP) technology, and the extracted keywords are matched according to a preset label library to generate corresponding task type labels; A regression algorithm is used to construct a task complexity prediction model, wherein the input parameters of the task complexity model comprise the number of code lines, the number of function modules and the number of dependency relationships, and the output parameter of the task complexity model is a task complexity score; The time cost is calculated according to the task complexity score, the historical completion time benchmark value of the personnel and the priority correction coefficient.

4. The intelligent dispatching method based on multi-dimensional project accounting according to claim 2, characterized in that, In S12, the personnel skill label is generated according to the personnel resume, training record and historical task skill usage statistics in the personnel description document; The historical performance quantitative value of the personnel is obtained, including: calculating the historical completion rate of the personnel on the task, generating a task quality score and a collaboration satisfaction degree; According to the remaining working hours of the personnel assigned task collected in real time, the load coefficient is generated by comparing the accumulated value with the upper limit of the load; The collaboration task data of the personnel with each collaboration role is recorded, the collaboration success rate is calculated, and the collaboration preference of the personnel is obtained according to the ranking of the collaboration success rate with each collaboration role.

5. The intelligent dispatching method based on multi-dimensional project accounting according to claim 1, characterized in that, In S2, the weights of the personnel dimension data of personnel in different roles are configured differently according to the role difference of the personnel.

6. The intelligent dispatching method based on multi-dimensional project accounting according to claim 1, characterized in that, In S2, the weights of each dimension are dynamically adjusted according to the real-time scene, including: (1) Based on a rule engine, the weights of the corresponding dimensions are dynamically adjusted according to preset adjustment rules, the preset adjustment rules including: a preset trigger condition and an adjustment rule of the weight of the task dimension and the weight of the personnel dimension corresponding to the trigger condition, wherein the trigger condition includes: a change in task state, a change in personnel state, or a change in both; (2) Based on feedback learning, the weights of the task dimension and the weight of the personnel dimension are automatically optimized through the correlation analysis of historical dispatching results and actual completion results.

7. The intelligent dispatching method based on multi-dimensional project accounting according to claim 1, characterized in that, In S4, the candidate personnel are screened, including: (1) Automatically exclude the candidate personnel with abnormal load coefficients; (2) Judge the matching degree score of the candidate personnel, and when the matching degree score is lower than the preset matching degree threshold, trigger manual intervention; (3) According to the role of the personnel initiating the dispatching, preferentially select personnel who are task-collaborative adaptive with the personnel initiating the dispatching. 8.The intelligent dispatching method based on multi-dimensional project accounting of claim 1, wherein, In S4, the scores of each dimension of the display task dimension and the personnel dimension are displayed, and a matching degree sorting table is displayed, the matching degree sorting table including: the comprehensive score of each candidate personnel and the dimension contribution value. 9.The intelligent dispatching method based on multi-dimensional project accounting of claim 1, wherein, When it is detected that the task dimension data or the personnel dimension data is missing, the completion operation of the missing dimension data is automatically triggered; When it is detected that the matching degree scores of all candidate personnel are lower than the preset matching degree threshold, an alarm is issued, and an optimization direction suggestion for the current dispatching task is automatically pushed; When it is detected that the personnel is overloaded, an alarm is issued, and a list of replacement personnel is provided.

10. The intelligent dispatching method based on multi-dimensional project accounting according to claim 1, characterized in that, When a network interruption is detected, data breakpoint resume is supported; when it is detected that there is a dimension with data anomaly, the dimension is automatically excluded.

Images