Scheduling optimization method and device

Abstract

The invention discloses a scheduling optimization method and device, and the method comprises the steps: inputting external dynamic data and internal scheduling data into a compliance risk factor model, and outputting values corresponding to different risk factors; performing standardization processing on values corresponding to the risk factors to obtain feature vectors, inputting the feature vectors to the man-hour compliance prediction model, and outputting compliance risk prediction results including quantitative prediction results of overtime overrun probability, on-duty compliance rate and policy adaptation deviation rate; a target scheme type is matched based on the compliance risk prediction result, the scheduling scheme is determined based on the target scheme type, and the scheduling scheme comprises a remote office scheme, a shift adjustment scheme, an elastic man-hour scheme and a substitute scheduling scheme. The compliance risk is predicted based on the multi-source data, and the scheduling scheme is intelligently determined based on the compliance risk, so that the defects of manual scheduling can be overcome.

Description

Technical Field

[0001] This invention relates to the field of information processing technology, specifically to a scheduling optimization method and apparatus. Background Technology

[0002] The core scheduling challenges in labor-intensive industries are their large workforce (ranging from hundreds to thousands), complex time management, high compliance risks, and low mismatch between manpower and business needs. They often face a dilemma: prioritizing compliance at the expense of business (e.g., reducing manpower to avoid overtime leading to order delays) or prioritizing business while neglecting compliance (e.g., penalties resulting from mandatory overtime). There is a lack of quantifiable balancing mechanisms. The core compliance challenges in labor-intensive industries are that manual scheduling easily leads to excessive overtime, violations of on-duty regulations (e.g., concentrated employee lateness), and policy misalignment (e.g., failure to implement regional employment policies).

[0003] Therefore, overcoming the aforementioned problems with manual scheduling in labor-intensive business sectors is an urgent issue that needs to be addressed. Summary of the Invention

[0004] The main objective of this invention is to provide a scheduling optimization method to address the shortcomings of related technologies.

[0005] To achieve the above objectives, according to a first aspect of the present invention, a scheduling optimization method is provided, comprising: inputting external dynamic data and internal scheduling data into a compliance risk factor model, and outputting values ​​corresponding to different risk factors; standardizing the values ​​corresponding to the risk factors to obtain feature vectors, and then inputting them into the working hour compliance prediction model, and outputting compliance risk prediction results, wherein the compliance risk prediction results include quantitative prediction results of overtime exceedance probability, on-duty compliance rate, and policy adaptation deviation rate; matching target scheme types based on the compliance risk prediction results, and determining the scheduling scheme based on the target scheme types, wherein the scheduling scheme includes remote work schemes, shift adjustment schemes, flexible working hour schemes, and substitute scheduling schemes.

[0006] According to a second aspect of the present invention, a scheduling optimization device is provided, comprising: a first prediction unit, configured to input external dynamic data and internal scheduling data into a compliance risk factor model, and output values ​​corresponding to different risk factors; a second prediction unit, configured to standardize the values ​​corresponding to the risk factors to obtain feature vectors, and input them into the working hour compliance prediction model, and output compliance risk prediction results, wherein the compliance risk prediction results include quantitative prediction results of overtime exceedance probability, on-duty compliance rate, and policy adaptation deviation rate; and a scheme generation unit, configured to match target scheme types based on the compliance risk prediction results, and determine the scheduling scheme based on the target scheme types, wherein the scheduling scheme includes remote work schemes, shift adjustment schemes, flexible working hour schemes, and substitute scheduling schemes.

[0007] According to a third aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing the computer to perform the method described in any one of the first aspects.

[0008] According to a fourth aspect of the present invention, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to cause the at least one processor to perform the method described in any implementation of the first aspect.

[0009] This embodiment of the scheduling optimization method and apparatus includes inputting external dynamic data and internal scheduling data into a compliance risk factor model, outputting values ​​corresponding to different risk factors; standardizing the values ​​corresponding to the risk factors to obtain feature vectors, then inputting them into the work hour compliance prediction model, outputting compliance risk prediction results, which include quantitative prediction results of overtime exceeding limits probability, on-duty compliance rate, and policy adaptation deviation rate; matching target scheme types based on the compliance risk prediction results, and determining the scheduling scheme based on the target scheme types, wherein the scheduling scheme includes remote work schemes, shift adjustment schemes, flexible working hour schemes, and substitute scheduling schemes. Predicting compliance risks based on multi-source data and intelligently determining scheduling schemes based on compliance risks can overcome the shortcomings of manual scheduling. Attached Figure Description

[0010] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0011] Figure 1 This is a flowchart of the scheduling optimization method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0012] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0013] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of the invention described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0014] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0015] According to embodiments of the present invention, a scheduling optimization method is provided, such as... Figure 1 As shown, steps 101 to 103 are included below: Step 101: Input external dynamic data and internal scheduling data into the compliance risk factor model and output the values ​​corresponding to different risk factors.

[0016] In this step, external dynamic data refers to environmental or policy data generated outside the enterprise that changes dynamically over time / scenario and may affect work hour compliance. It is characterized by strong timeliness and uncontrollability, including meteorological data such as weather type (heavy rain / normal), warning level (red / orange / yellow / blue), wind force level (0-12), precipitation (mm / 12 hours), etc.; traffic data such as commuter route congestion index (0-10), controlled road section coverage rate (0-100%), estimated delay time (minutes), etc.; policy data such as public health policies (such as work-from-home requirements), labor law updates (such as local work hour adjustments), policy scope (all regions / partial areas), etc.

[0017] Internal scheduling data refers to the basic data generated within an enterprise that is related to employee scheduling and work hour management. It is manageable and structured, and specifically includes basic employee data such as job type (core position / general position), skill tags, historical attendance compliance rate (0-1), monthly cumulative overtime hours (hours), etc.; scheduling plan data: daily / next day schedule (including arrival / departure time), manpower requirements for each time period (e.g., 10 people needed for the morning peak), core business hours (e.g., 9:00-18:00), etc.; compliance benchmark data such as legal working hour thresholds (e.g., daily overtime ≤ 3 hours, monthly ≤ 36 hours), and internal working hour rules of the enterprise (e.g., flexible working hour deviation range), etc.

[0018] The compliance risk factor model transforms external dynamic data and internal scheduling data into a model that quantifies risk factors. Its core function is to map input data into standardized values ​​that characterize risk triggers.

[0019] Risk factors refer to quantitative indicators (numericalized) that characterize various compliance risk triggers and are used for subsequent compliance risk prediction. These include external dynamic risk factors, generated based on external dynamic data, such as weather impact factors (0-1), commuting interference factors (0-1), and policy constraint factors (0-3); and internal basic risk factors, generated based on internal scheduling data, such as employee working hour compliance rate (0-1), job vacancy rate (0-1), and current cumulative overtime hours (hours, associated with statutory thresholds).

[0020] Before processing, the model preprocesses the input data, including data cleaning. During data cleaning, outliers are removed (e.g., traffic congestion index = -1, which is marked as invalid and filled with the average of the first 3 times); missing values ​​are filled in.

[0021] This also includes data standardization. During standardization, for external dynamic data, different levels of indicators are mapped to the 0-1 or 0-3 range (e.g., warning levels "red = 4, orange = 3, yellow = 2, blue = 1" are normalized to 0-1: red = 1.0, orange = 0.75, yellow = 0.5, blue = 0.25; policy constraint factors are directly quantified as "no policy = 0, partial flow restriction = 1, partial home stay = 2, full area home stay = 3"); internal scheduling data is converted from absolute values ​​to relative proportions (e.g., current cumulative overtime hours = 25 hours, statutory monthly threshold = 36 hours are converted to "cumulative overtime percentage = 25 / 36≈0.69"; vacancy rate = (number of people needed - number of people on duty) / number of people needed).

[0022] It also includes spatiotemporal alignment, unifying all data to hourly time granularity during time alignment (e.g., the "future 12 hours" of weather warnings and the "8:00-20:00" of the work schedule correspond to hourly slices); and spatial alignment: matching external data according to employee commuting area and job location area (e.g., only including traffic control data covered by employee commuting routes in the calculation).

[0023] The compliance risk factor model transforms preprocessed data into risk factors, including: Calculation of external dynamic risk factors, such as weather impact factor (0-1): Weather impact factor = Normalized value of warning level × 0.6 + Correction value for accompanying conditions × 0.4 Example: A red rainstorm warning (normalized to 1.0) + wind force 8 (with accompanying condition correction value 0.8) is 1.0 × 0.6 + 0.8 × 0.4 = 0.92; the accompanying condition correction values ​​are 0.8-1.0 for wind force ≥ 8 / precipitation ≥ 100mm, 0.4-0.7 for wind force 3-7 / precipitation 25-100mm, and 0.1-0.3 for precipitation < 3 / precipitation < 25mm.

[0024] For example, the commuting interference factor (0-1) is calculated as follows: Commuting interference factor = Congestion index / 10 × 0.5 + Controlled road section coverage ratio × 0.5.

[0025] Example: Congestion index 8 (8 / 10=0.8) + 60% of employee commuting routes covered by regulations, i.e. 0.8×0.5 + 0.6×0.5=0.7.

[0026] For example, when calculating the policy constraint factor (0-3), the preset level can be directly matched according to the policy type (e.g., 3 for working from home across the entire region and 1 for localized flow restriction), and the policy text keyword weight can be adjusted (including the keyword "administrative penalty" +0.5, with a maximum of 3).

[0027] This also includes calculations of internal basic risk factors, such as employee work hour compliance rate (0-1), with the algorithm formula: Work hour compliance rate = Sum of compliant work hours in the past 30 days / Total work hours in the past 30 days; compliant work hours refer to work hours that meet the requirement of "≤8 hours per day + ≤3 hours of overtime", and the excess is not included in compliant work hours. Job vacancy rate (0-1), job vacancy rate = Σ(Number of employees needed in each time period - Number of employees on duty) / Σ Number of employees needed in each time period (only core business time periods are calculated); Current cumulative overtime hours (hours): "Employee monthly cumulative overtime hours" are directly extracted from internal scheduling data and associated with a statutory monthly threshold (e.g., 36 hours) for subsequent parameter calculations.

[0028] Furthermore, the model outputs a risk factor data table with fields including: factor name, value, calculation time, and data source index (e.g., "Weather Influence Factor = 0.92, Source: Meteorological API 2025-10-15 08:00 Data"). Furthermore, when business scenarios or the external environment change (such as sudden public policy changes or the launch of new businesses), the original weights may become invalid, automatically triggering weight adjustments to ensure the timeliness of the weights. Triggering conditions and adjustment logic include changes in the influence patterns of new data dimensions (such as "sudden public event level") or existing dimensions (such as "a sudden decrease in the impact of the congestion index in commuting interference factors"); policy triggers, such as the release of new working hour regulations (such as shortening the daily overtime threshold), require adjustments to compliance-related weights (such as increasing the weight of dimensions related to "statutory working hour threshold"). In specific adjustments, small adjustments can be made, such as fine-tuning the current optimal weights by a specified margin to determine the new weights. Each weight adjustment record is automatically saved, including the adjustment time, triggering conditions, and the old and new weight values, supporting backtracking and auditing.

[0029] Step 102: After standardizing the values ​​corresponding to the risk factors to obtain feature vectors, input them into the working hour compliance prediction model and output the compliance risk prediction results. The compliance prediction model is trained using the gradient boosting tree algorithm with historical compliance risk events as labels. The compliance risk prediction results include quantitative prediction results of overtime exceedance probability, on-duty compliance rate, and policy adaptation deviation rate.

[0030] In this step, historical data is used to teach the gradient boosting tree algorithm how risk factors affect outcomes such as overtime exceeding limits and on-duty compliance. A multi-output structure is used to simultaneously predict three types of risks. During the online prediction phase, real-time risk factors are standardized into feature vectors, which are then input into the model to directly output quantitative prediction results. This provides accurate risk information for subsequent scheduling plans generated based on the prediction results.

[0031] Step 103: Match the target scheme type based on the compliance risk prediction results, and determine the scheduling scheme based on the target scheme type, wherein the scheduling scheme includes remote work scheme, shift adjustment scheme, flexible working hours scheme, and substitute scheduling scheme.

[0032] In this step, the executable plan generated based on the target solution type includes core parameters (such as remote coverage ratio and shift delay duration), employee execution details (such as remote employee list and adjusted arrival time), and compliance verification items (such as whether it complies with legal working hours). It is a structured document that connects the solution type with the actual shift schedule.

[0033] Matching target solution types based on compliance risk prediction results aims to transform scheduling solutions from unfounded and blind adjustments to precise interventions, avoiding problems such as the disconnect between solutions and risks, waste of resources, and imbalance between compliance and business in traditional scheduling.

[0034] The compliance risk prediction results (such as a probability of excessive overtime (0.85%), an on-duty compliance rate (0.58%), and a policy adaptation deviation rate (0.72%) clearly identify the current / future core risk types (whether it's overtime, on-duty compliance, or policy issues) and their severity. Matching the solution type based on this allows the solution to "directly address the pain points" rather than making general adjustments. The compliance risk prediction results not only include quantitative values ​​of compliance risk but also implicitly reveal the relationship between risk and business operations. For example, a high risk of excessive overtime may be due to insufficient manpower during peak business periods. Matching the solution type based on this allows for a balance between reducing compliance risk and ensuring business operations.

[0035] It overcomes the problems of slow response and logical disconnect in the traditional process, where compliance risk prediction and scheduling plan generation rely heavily on manual coordination.

[0036] As an optional implementation method in this embodiment, matching target scheme types based on compliance risk prediction results includes: matching different types of target schemes based on the magnitude relationship between quantitative prediction results and preset quantitative thresholds.

[0037] In this optional implementation, when matching the target solution type with the quantitative value based on the compliance risk prediction result, the quantitative value threshold is used as the judgment standard to construct a quantitative value and solution mapping rule base. The rule base takes the quantitative value threshold of three types of risks (policy adaptation deviation rate, on-duty compliance rate, and overtime exceedance probability) as the core condition, clarifies the target solution type corresponding to different threshold ranges, and defines the priority of compound risks (multiple types of quantitative value exceedance) to ensure that the matching logic is unambiguous.

[0038] For example, regarding the policy adaptation deviation rate P1, if P1 ≥ K1, the deviation rate is high, and a remote work plan is matched; if K1 ≥ P1 ≥ K2, the deviation rate is moderate, and a remote work plan is also matched. Regarding the on-site compliance rate P2, if P2 ≤ K3, the compliance rate is low, and a shift adjustment plan is matched; if K3 ≤ P2 ≤ K4, a shift adjustment plan is also matched. Regarding the probability of exceeding overtime limits P3, if P3 ≥ K5, the probability is high, and a flexible working hours plan is matched; if K5 ≥ P3 ≥ K6, the probability is moderate, and a flexible working hours plan is matched. If P1 < K2, P2 > K4, and P3 < K6, the existing shift schedule is maintained. The above critical thresholds can be adjusted as needed.

[0039] During implementation, the specific values ​​of the three types of risks are read from the compliance risk prediction results (e.g., "policy adaptation deviation rate 0.72, on-site compliance rate 0.58, and overtime exceeding probability 0.85"). If only one type of risk has a quantified value exceeding the low-risk range (e.g., only the policy adaptation deviation rate 0.72 ≥ 0.6), the corresponding solution type (remote work solution) is directly matched. If multiple types of risks have quantified values ​​exceeding the low-risk range (e.g., policy adaptation deviation rate 0.72 ≥ 0.6 + on-site compliance rate 0.58 ≤ 0.6), the main solution and auxiliary solution are determined according to the preset priority (policy adaptation deviation > on-site compliance > overtime exceeding probability). Primary Solution: Prioritizes solutions corresponding to high-priority risks (e.g., if the policy adaptation deviation rate is high, the primary solution is the remote work solution); Secondary Solution: Supplements solutions corresponding to low-priority risks (e.g., if the on-site compliance rate is low, the secondary solution is the shift adjustment solution). Example: If the policy adaptation deviation rate is 0.72 + the on-site compliance rate is 0.58, the matching result is that the primary solution is the remote work solution, and the secondary solution is the shift adjustment solution.

[0040] The essence of calculating the core parameters of the scheme is to transform the abstract scheme type into specific operational instructions that can be implemented, comply with regulations, and adapt to business needs. This is a key link to ensure that the scheduling scheme can accurately address compliance risks without affecting business operations.

[0041] As an optional implementation of this embodiment, a specified parameter is calculated based on the compliance risk prediction result and the value of the dynamic risk factor. For different target scheme types, the compliance risk prediction result is converted into a base value and the value corresponding to different risk factors is converted into a modified value using preset rules. The parameter is determined based on the base value and the modified value.

[0042] In this optional implementation, a base value calculation rule and a correction value calculation rule are preset for each scheduling scheme type. The base value calculation rule takes only the compliance risk prediction results (overtime exceedance probability, on-duty compliance rate, policy adaptation deviation rate) as input and outputs the initial parameter values ​​through a fixed formula. The correction value calculation rule takes only dynamic risk factors (external: weather / commuting / policy factors; internal: working hour compliance rate / absence rate / cumulative overtime hours / statutory threshold) as input and outputs the correction amount through coefficient adjustment.

[0043] Furthermore, parameter calculations are performed according to the scheme type, and the execution logic is as follows: a) Remote work solution, the core parameters of which include remote coverage ratio and elastic time supplement limit.

[0044] Calculate the base value: Remote coverage ratio base value = Policy adaptation deviation rate × 100%; Flexible time supplement limit base value = Statutory working hours threshold (daily overtime limit) × 0.3; Calculate the correction value: Coverage ratio correction value = (Policy constraint factor / 3) × 20% - (Job vacancy rate × 10%); Time supplement limit correction value = Employee working hours compliance rate × 0.1; Obtain the final value: Coverage ratio = Base value + Correction value (80% + 20% - 1% = 99%); Time supplement limit = Base value + Correction value (0.9 + 0.09 = 0.99 hours).

[0045] b) Core parameters of the schedule adjustment plan: Delay duration of the schedule: Calculate the base value: Delay duration base value = (1 - on-duty compliance rate) × 2; Calculate the correction value: Delay duration correction value = Commuting interference factor × 0.5 + Weather impact factor × 0.5; Obtain the final value: Delay duration = Base value + Correction value (1 + 0.85 = 1.85 hours). c) The core parameter of the flexible working hours scheme is the flexible deviation range.

[0046] Calculate the base value: Deviation range base value = Overtime exceedance probability × 1.5; Calculate the correction value: Deviation range correction coefficient = [1 - (Current cumulative overtime hours / Statutory monthly threshold)] × Employee working hour compliance rate; Obtain the final value: Deviation range = Base value × Correction coefficient; d) Core parameter of the substitute scheduling scheme: Number of substitutes: Calculate the base value: Base value of the number of substitutes = rounded up (probability of exceeding overtime limits × job vacancy rate × 50); Calculate the correction value: Correction value of the number of substitutes = rounded down (employee working hour compliance rate × 2) (compliance 0.9 → 1 person); Obtain the final value: Number of substitutes = Base value + Correction value.

[0047] Regarding the mapping between parameters and rules, a pre-set rule base is used to establish a connection between core parameters and scheduling actions. The rule base includes the mapping relationship between parameter types and scheduling elements (such as the mapping between remote coverage ratio and the range of employees who need to work remotely, and the mapping between flexible deviation range and the time interval that employees can adjust). When binding core parameters to specific objects, employee attribute data and business requirement data are invoked to bind the core parameters to specific objects (employees, positions, time periods) (e.g., selecting qualified personnel from employee data based on "remote coverage ratio" and matching the manpower requirements of corresponding positions based on "core business time periods"). During the construction of structured solutions, based on the mapping relationship and binding results, a structured solution containing individual employee scheduling details, departmental time period manpower distribution, and compliance verification items is automatically generated. The solution undergoes compliance (compliance with legal working hour standards) and business adaptability (manpower requirements for core time periods) checks. If a check fails, fine-tuning within thresholds is performed based on the core parameters (e.g., adjusting the number of substitutes or compressing the flexibility range).

[0048] Furthermore, when generating a structured scheme based on the mapping relationship and the binding result, an automated process of rule parsing, data matching, detail generation, summary verification, and structured encapsulation is used to transform the abstract mapping relationship and the specific binding result into an executable scheduling scheme. The specific implementation logic is as follows: Based on the correspondence rules between core parameters and scheduling elements in the mapping relationship, suitable employees are selected from the employee attribute data of the binding results (e.g., remote coverage ratio corresponds to selecting employees with "job digitization rate ≥ parameter threshold", shift delay duration corresponds to matching "employees who need to adjust their arrival time"); specific scheduling element values ​​are assigned to the selected employees, generating individual detailed records containing "employee ID, department, arrival time, departure time, arrival type (remote / on-site), flexible time period range (if it is a flexible working hour plan), and substitute flag (if it is a substitute scheduling plan)"; further, the individual employee scheduling details are aggregated by department and time period dimensions, and the whole day is divided into fixed time slices (every 30 minutes / 1 hour), and the actual number of employees on duty in each department in each time slice (including on-site employees + remote core time period employees) can be counted, and finally, the department time period manpower distribution can be generated; further, the "required number of people" of each department in each time period is extracted from the business time period demand data of the binding results, generating a record containing "department, time period slice, actual number of people on duty, required number of people, manpower gap / surplus (actual number of people - The system generates a distribution table for "Number of employees required"; extracts "Daily working hours (departure time - arrival time), overtime hours (exceeding the daily working hour limit), and remote working hours percentage" from individual employee scheduling details; extracts "Core time period on-duty rate (actual number of employees in the core time period / number of employees required)" from departmental time period manpower distribution; retrieves compliance thresholds from the legal working hour standard data of the binding results, compares the extracted parameters with the thresholds, and generates a verification list containing "Verification ID, associated employee / department, verification parameter name, actual parameter value, legal threshold, compliance status (compliant / non-compliant), and rectification suggestions (such as shortening overtime hours if non-compliant)"; and integrates the results of each step to generate a structured scheduling plan.

[0049] Furthermore, the unified solution format is a structured format (JSON / Excel) compatible with the scheduling management system, adding associated indexes (employee ID, department ID, time period code) for individual employee scheduling details, departmental time period manpower distribution, and compliance verification items.

[0050] As an optional implementation of this embodiment, after the compliance risk factor model outputs the value corresponding to the risk factor, the risk factor coupling degree is calculated, wherein the risk factor coupling degree is a quantitative value of the correlation strength between the external dynamic risk factor and the internal basic risk factor; the coupling pair is determined based on the risk factor coupling degree, wherein the standardization of the value corresponding to the risk factor to obtain the feature vector includes standardizing the coupling pair into a feature vector.

[0051] External dynamic risk factors refer to quantitative indicators of risk factors generated by the external environment of an enterprise, which change in real time over time / scenario and are not directly controlled by the enterprise. Their core function is to reflect the degree of interference of the external environment on working hour compliance. Specifically, they include: weather impact factor (0-1), commuting interference factor (0-1), and policy constraint factor (0-3). The core characteristics of these factors are that they are highly dynamic and the data needs to be synchronized in real time (such as updating commuting data every 15 minutes), and the enterprise cannot directly change them through internal management (such as being unable to control rainstorms or traffic congestion).

[0052] Internal basic risk factors refer to quantitative indicators of risk triggers generated by internal management, reflecting the internal operational status, and directly controllable by the enterprise. Their core function is to reflect the degree to which internal resources / management support work hour compliance. Specifically, they include: employee work hour compliance rate, job vacancy rate (0-1), and current cumulative overtime hours. The core characteristic of these factors is their strong structure; data is stored in the enterprise's internal systems (HR, scheduling systems).

[0053] When calculating the Pearson correlation coefficient for each pair of risk factors using historical data, the pairings are not randomized. Factors that may jointly affect a certain type of compliance risk are paired to ensure that the calculated coupling (correlation coefficient) reflects the true risk synergy. The specific pairing logic is divided into three categories, as follows: 1. Category 1: Pairs of external dynamic risk factors (synergistic effects within the external environment) The full combination of external factors, due to the frequent synchronous changes in the external environment (e.g., heavy rain is often accompanied by traffic congestion), the specific combination and correlation logic are as follows: Pair 1: The correlation between weather impact factor and commuting interference factor: Severe weather (such as heavy rain and heavy snow) will directly aggravate commuting congestion. Both factors together affect the "on-duty compliance rate" (for example, during heavy rain, the weather factor is 0.9 + the commuting factor is 0.8, and the on-duty compliance rate drops significantly). Pair 2: The correlation between weather influencing factors and policy constraint factors: Extreme weather (such as typhoons) may trigger temporary policies (such as work stoppage notices), and the two together affect the "policy adaptation deviation rate" (such as typhoon days + work stoppage policy, the deviation rate of not implementing work stoppage needs to be judged in combination with the intensity of weather and policy). Pair 3: The correlation between commuting interference factors and policy constraint factors: Regional traffic control policies (such as traffic restrictions) will directly increase commuting interference, and the two together affect the "competition rate of arriving at work" (such as traffic restriction policy + commuting congestion, the probability of arriving at work is higher).

[0054] 2. Second category: Pairing of internal basic risk factors. The pairing method is targeted combination of internal factors, only combining factors that are strongly correlated with the same compliance risk. Specific combination and correlation logic: Pair 1: Job vacancy rate ↔ Current cumulative overtime hours correlation logic: Job vacancy (e.g., 2 vacancies in core positions) will lead to an increase in overtime for existing employees. The two together affect the "probability of exceeding overtime limits" (e.g., if the vacancy rate is 0.2 + cumulative overtime hours of 25 hours, the probability of exceeding the limit is significantly higher than the impact of a single factor). Pair 2: Employee working hours compliance rate ↔ legal working hours threshold correlation logic: When an employee's historical compliance rate is low (e.g., 0.7) and the legal threshold is strict (e.g., 2 hours of overtime per day), it is easier to trigger overtime exceeding the limit. The two together affect the "probability of overtime exceeding the limit". Pair 3: The correlation between job vacancy rate and employee working hour compliance rate: Job vacancy may lead to a decrease in employee compliance rate (long-term overtime leads to fatigue), and the two together affect the "on-duty compliance rate" (e.g., if the vacancy rate is 0.3 + compliance rate is 0.6, the probability of being late for work is higher).

[0055] 3. The third category: Pairs between external dynamic risk factors and internal fundamental risk factors. The pairing method involves cross-category associations between external and internal factors, combining only factors that have interactive effects. Specific combination and association logic are as follows: Pair 1: The correlation between commuting disruption factor and employee work hour compliance rate: When commuting disruption is significant (e.g., 1 hour of traffic congestion), if an employee's historical compliance rate is low (e.g., 0.6), they are more likely to be late for work. Both factors together affect the "on-duty compliance rate". Pair 2: The correlation between weather impact factors and job vacancy rates: Severe weather (such as heavy rain) may cause some employees to take leave, exacerbating job vacancy. Both factors together affect the "probability of working overtime beyond the limit" (e.g., weather factor 0.8 + vacancy rate 0.2, the overtime pressure on existing employees will double). Pair 3: The correlation between policy constraint factors and job vacancy rate: If remote work policies (such as policy factor 3) are combined with job vacancy (such as a vacancy of 1 person for a remote position), it will lead to insufficient remote work coverage. The two together affect the "policy adaptation deviation rate".

[0056] Active coupling pairs are screened after pairing: After completing all associated pairings, not all pairs participate in subsequent corrections - only pairs with "absolute Pearson correlation coefficient ≥ 0.6" are retained, and weakly associated pairs (such as policy constraint factors and employee working hour compliance rate, the correlation coefficient is usually < 0.3, with no synergistic effect) are excluded, to ensure that subsequent risk factor corrections focus on real synergistic effects and avoid introducing invalid interference.

[0057] Examples of coupling pairs are as follows: External-External coupling pair: Weather impact factor (W) × Commuting interference factor (T) (correlation coefficient 0.8, strong positive correlation); Internal-Internal coupling pair: Job vacancy rate (S) × Current cumulative overtime hours (O) (correlation coefficient 0.7, strong positive correlation); External-Internal coupling pair: Policy constraint factor (P) × Job vacancy rate (S) (correlation coefficient -0.65, strong negative correlation, the impact of vacancy is weakened when policy constraints are strong).

[0058] After identifying the coupling pairs, coupling feature terms are generated by multiplying the factor values. The formula is: Coupling feature term = Standardized value of factor A × Standardized value of factor B. For example: Standardized value of weather impact factor 0.9, standardized value of commuting interference factor 0.8 → Coupling feature term = 0.9 × 0.8 = 0.72; For example: Standardized value of job vacancy rate 0.3, standardized value of cumulative overtime hours 0.6 → Coupling feature term = 0.3 × 0.6 = 0.18.

[0059] As an optional implementation of this embodiment, the compliance prediction model is trained using a gradient boosting tree algorithm with historical compliance risk events as labels. The training method for the compliance prediction model includes: determining coupled pairs of samples based on historical data; generating coupled feature terms for each coupled pair of samples; appending the generated coupled feature terms to the original risk factor feature vector in a fixed order to form an extended feature vector; and inputting all the extended feature vectors into the compliance prediction model.

[0060] In this optional implementation, after processing historical samples, coupling pairs are generated and then fused with the original risk factor features to form an expanded feature matrix. This includes retaining the standardized original risk factor feature vectors; and appending coupling feature terms, which involves appending the generated coupling feature terms to the original feature vectors in a fixed order to form an expanded feature vector. The above operations are performed on all historical training samples, combining the original feature vector and coupling feature terms of each sample to form an expanded historical feature matrix (number of rows = number of samples, number of columns = number of original features + number of coupling feature terms), which serves as the input for model training.

[0061] In this optional implementation, multiple decision trees are trained iteratively, with each tree focusing on correcting the prediction errors of the preceding trees. The outputs of all trees are then weighted and fused to improve prediction accuracy. Its core advantage is its ability to capture the non-linear relationship between risk factors and compliance risks (such as the abrupt impact of a sudden increase in commuting disruption factors on on-duty compliance rates), making it suitable for processing structured risk factor data.

[0062] Historical compliance risk events refer to specific events related to working hours compliance that actually occurred in the company's historical records. They serve as the basis for the actual results of model training and include three categories: overtime exceeding limits events, such as events where employees' actual overtime hours exceed the legal threshold (e.g., employee A works 4 hours of overtime in a single day, exceeding the legal threshold of 3 hours); on-duty violation events, such as events where employees do not arrive at work according to the scheduled time (e.g., employee B is supposed to arrive at 8:00, but actually arrives at 9:30); and policy adaptation deviation events, such as events where the schedule is not implemented according to policy requirements (e.g., policy requires working from home in the region, but employee C still arrives at work).

[0063] The training process of the compliance prediction model is essentially to enable the gradient boosting tree algorithm to learn the correlation between historical risk factors and their corresponding labels, thereby mastering the ability to predict compliance risks from risk factors.

[0064] The training process is as follows: Compared to the original features, the expanded features have increased feature dimensions. Therefore, the number of decision trees should be increased appropriately (e.g., from 300 to 400), while keeping the depth of each tree (3-8 layers) unchanged to avoid overfitting. Feature weight initialization: The coupled feature terms are assigned the same initial weights as the original features (e.g., both are 1.0) to ensure that the algorithm learns the influence of all features fairly.

[0065] Decision Tree Splitting and Coupled Feature Learning: The core of gradient boosting trees is to construct a decision tree through feature splitting, with coupled feature terms participating in the splitting as independent features. The specific process is as follows: Training the first tree: The algorithm traverses all features (including coupled feature terms) and calculates the information gain (the reduction in error after splitting) of each feature on the label (such as the probability of working overtime beyond the limit). Example: If the information gain (0.35) of “W×T” is higher than that of a single feature W (0.2) or T (0.25), then “W×T” is preferred as the splitting node (e.g., when W×T≥0.6, the sample enters the left subtree, and the probability of predicting overtime exceeding the limit is high). Subsequent tree training: Each new tree focuses on correcting the prediction error of the previous tree. If the error mainly stems from "insufficient consideration of synergistic effects" (e.g., for a sample W=0.5, T=0.5, the single feature has a low impact, but the coupled feature W×T=0.25 still causes prediction bias), then the new tree will prioritize splitting based on the coupled feature terms to strengthen the correction of synergistic effects.

[0066] Error Calculation and Iterative Optimization: Prediction error includes "single factor influence error" and "co-influence error": After introducing coupled feature terms, the model can directly reduce the co-influence error by adjusting the weights of the coupled feature terms (such as increasing the weight of W×T); Iteration Termination Condition: When the prediction error (such as mean squared error MSE) of the model trained with the extended feature matrix on the validation set is reduced by ≥15% compared to the "model without coupled feature terms", and no longer decreases for 5 consecutive rounds, training is stopped to ensure that the coupled feature terms effectively improve accuracy.

[0067] Model parameter storage: After training, the model parameters include not only the splitting threshold and weight of the original features, but also the splitting threshold (e.g., W×T≥0.6) and weight (e.g., in a certain tree, the influence weight of W×T is 0.3), ensuring that the learning results of the coupled features can be directly called during online prediction.

[0068] In this optional implementation, the essence of introducing coupled feature terms is to transform the implicit synergistic effects between risk factors into explicit features, allowing the gradient boosting tree algorithm to directly learn these effects. During the generation phase, the synergistic effect of "W and T jointly exacerbating the risk of on-duty attendance" is transformed into a quantifiable "W×T" feature through the product of active coupled pairs. During the introduction phase, coupled features are fused with original features to expand the model's input dimension, providing a data foundation for learning synergistic effects. During the training phase, the gradient boosting tree automatically strengthens the learning of synergistic effects by prioritizing the splitting of coupled feature terms with higher information gain. Ultimately, this allows the model to capture both the independent effects of a single factor and the combined effects between factors, significantly improving prediction accuracy in complex scenarios (such as predicting on-duty compliance rates when heavy rain and traffic congestion occur simultaneously).

[0069] This implementation overcomes the shortcomings of existing scheduling models, which often handle single risk factors independently (e.g., only considering weather or only considering commuting) and neglect the synergistic effects between factors (e.g., the combined impact of heavy rain and traffic congestion on on-site compliance rate). This leads to significant deviations in compliance risk prediction. For example, implicit correlations between risk factors (e.g., increased policy constraints weaken the impact of staff shortages on overtime) are not quantified, and relying solely on human experience can easily result in "over-adjustment" or "under-adjustment." Furthermore, the lack of modeling capabilities for "coupled risks" results in low prediction accuracy (typically ≤70%) for overtime exceeding limits and on-site violations.

[0070] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0071] According to an embodiment of the present invention, a scheduling optimization device is also provided, comprising: a first prediction unit, used to input external dynamic data and internal scheduling data into a compliance risk factor model, and output values ​​corresponding to different risk factors; a second prediction unit, used to standardize the values ​​corresponding to the risk factors to obtain feature vectors, and input them into the working hour compliance prediction model, and output compliance risk prediction results, the compliance risk prediction results including quantitative prediction results of overtime exceedance probability, on-duty compliance rate, and policy adaptation deviation rate; and a scheme generation unit, used to match target scheme types based on the compliance risk prediction results, and determine the scheduling scheme based on the target scheme types, wherein the scheduling scheme includes remote work scheme, shift adjustment scheme, flexible working hour scheme, and substitute scheduling scheme.

[0072] According to embodiments of the present invention, the present invention also provides an electronic device, the electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to implement the methods described in any of the above embodiments.

[0073] According to embodiments of the present invention, the present invention also provides a readable storage medium storing computer instructions that enable a computer to perform the methods described in any of the above embodiments when executed.

[0074] According to embodiments of the present invention, the present invention also provides a computer program product that, when executed by a processor, can implement the methods described in any of the above embodiments.

[0075] Figure 2A schematic block diagram of an example electronic device 300 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices.

[0076] like Figure 2 As shown, the electronic device 300 includes a computing unit 301, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 302 or a computer program loaded from a storage unit 308 into a random access memory (RAM) 303. The RAM 303 may also store various programs and data required for the operation of the electronic device 300. The computing unit 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0077] Multiple components in electronic device 300 are connected to I / O interface 305, including: input unit 306, such as keyboard, mouse, etc.; output unit 307, such as various types of displays, speakers, etc.; storage unit 308, such as disk, optical disk, etc.; and communication unit 309, such as network card, modem, wireless transceiver, etc. Communication unit 309 allows electronic device 300 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0078] The computing unit 301 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 301 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 301 performs the various methods and processes described above, such as the object matching method. For example, in some embodiments, the object matching method may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 308. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 300 via ROM 302 and / or communication unit 309. When the computer program is loaded into RAM 303 and executed by the computing unit 301, one or more steps of the methods described above may be performed.

[0079] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0080] The program code used to implement the methods of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0081] In the context of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

Claims

1. A scheduling optimization method, characterized in that, include: External dynamic data and internal scheduling data are input into the compliance risk factor model, and the values ​​corresponding to different risk factors are output. After standardizing the values ​​corresponding to the risk factors to obtain feature vectors, the vectors are input into the working hour compliance prediction model to output compliance risk prediction results. The compliance risk prediction results include quantitative prediction results of overtime exceedance probability, on-duty compliance rate, and policy adaptation deviation rate. The target scheme type is matched based on the compliance risk prediction results, and the scheduling scheme is determined based on the target scheme type. The scheduling scheme includes remote work scheme, shift adjustment scheme, flexible working hours scheme, and substitute scheduling scheme.

2. The scheduling optimization method according to claim 1, characterized in that, Matching target solutions based on compliance risk prediction results includes: matching different types of target solutions based on the quantitative prediction results and their relationship with preset quantitative thresholds.

3. The scheduling optimization method according to claim 2, characterized in that, Determining the scheduling scheme based on the target scheme type includes: loading the corresponding parameter calculation rule engine according to the target scheme type, so that the rule engine can calculate the specified parameters based on the compliance risk prediction results and the values ​​of dynamic risk factors; The core parameters are associated with scheduling actions by establishing a pre-defined rule base, wherein the rule base contains mapping relationships between different types of parameters and scheduling elements; It retrieves employee attribute data and business requirement data, and binds core parameters to specific objects; Based on the mapping relationship and the binding results, a structured solution containing individual employee shift details, departmental time period manpower distribution, and compliance verification items is automatically generated.

4. The scheduling optimization method according to claim 3, characterized in that, Based on the compliance risk prediction results and the values ​​of dynamic risk factors, the specified parameters are calculated for different target scheme types. Pre-defined rules are used to convert the compliance risk prediction results into basic values ​​and the values ​​corresponding to different risk factors into modified values. The parameters are determined based on the base value and the correction value.

5. The scheduling optimization method according to claim 1, characterized in that, After the compliance risk factor model outputs the corresponding values ​​of the risk factors, the risk factor coupling degree is calculated, wherein the risk factor coupling degree is a quantitative value of the correlation strength between external dynamic risk factors and internal basic risk factors. Determining coupling pairs based on the coupling degree of the risk factors, wherein standardizing the values ​​corresponding to the risk factors to obtain feature vectors includes standardizing the coupling pairs into feature vectors.

6. The scheduling optimization method according to claim 5, characterized in that, The compliance prediction model is trained using a gradient boosting tree algorithm, with historical compliance risk events as labels. The training method for the compliance prediction model includes: Determine coupling pairs based on historical data; For each coupled sample pair, generate a coupling feature term; The generated coupled feature terms are appended to the original risk factor feature vector in a fixed order to form an extended feature vector; All extended feature vectors are used as inputs to the compliance prediction model.

7. A scheduling optimization device, characterized in that, include: The first prediction unit is used to input external dynamic data and internal scheduling data into the compliance risk factor model and output the values ​​corresponding to different risk factors. The second prediction unit standardizes the values ​​corresponding to the risk factors to obtain feature vectors, then inputs them into the working hour compliance prediction model and outputs compliance risk prediction results. The compliance risk prediction results include quantitative prediction results of overtime exceedance probability, on-duty compliance rate, and policy adaptation deviation rate. The scheme generation unit is used to match the target scheme type based on the compliance risk prediction results, and determine the scheduling scheme based on the target scheme type. The scheduling scheme includes remote work scheme, shift adjustment scheme, flexible working hours scheme, and substitute scheduling scheme.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to perform the method according to any one of claims 1-6.

9. An electronic device, characterized in that, include: At least one processor; And a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, the computer program being executed by the at least one processor to cause the at least one processor to perform the method according to any one of claims 1-6.

10. A computer program product, characterized in that, When executed by a processor, the computer program implements the method described in any one of claims 1-6.

Images