“dynamic automation method and algorithm for optimizing (human, software, hardware) resources in it service development projects”

Abstract

The present invention relates to a technology for optimizing human, software (SW), and hardware (HW) resources generated in IT service development projects through dynamic automation. Specifically, the invention provides an integrated process in which (1) a machine learning prediction module (100) calculates future resource demand based on project management tools, server / network logs, SW license usage history, etc., (2) a decision engine (200) determines resource allocation logic, such as personnel reallocation, license increase / decrease, and server expansion / reduction, by referring to rule tables and priorities, (3) a resource control module (300) executes allocation / release commands for actual resources (personnel, licenses, servers), and (4) a monitoring module (400) collects and analyzes the above results and real-time project status to continuously correct prediction errors and decision logic. This enables the resolution of recurring issues during IT service development projects, such as manpower bottlenecks, license overloads or shortages, and server resource overuse or insufficient resources, effectively mitigating project schedule delays and cost increases. Furthermore, by continuously correcting discrepancies between the machine learning model and actual data during project execution through online or adaptive learning methods, it improves long-term accuracy and efficiently performs dynamic resource optimization. Consequently, various benefits are achieved, including shortened project schedules, cost reductions, and improved resource utilization. At the same time, managers can reduce their workload through automated resource allocation processes and further enhance project quality and stability.

Description

Technology Field

[0001] The present invention relates to the field of dynamically automating and optimizing human, software (SW), and hardware (HW) resources during the execution of IT service development projects. More specifically, it belongs to the fields of project management (PM) and cloud computing, which resolve project bottlenecks and inefficient resource allocation issues by predicting complex data such as project schedules, work issues, server load, and license usage based on machine learning (ML), calculating an optimal resource allocation policy by integrating rule tables and priorities in a decision engine, and then automatically executing and managing this policy through resource control modules and monitoring modules.

[0002] In other words, this invention relates to a technology that enables project schedule reduction, cost reduction, and further improvement in project quality by combining machine learning prediction algorithms and automated decision logic to efficiently reallocate and increase or decrease resources in real-time or periodically, thereby resolving issues such as over- or under-deployment of personnel, waste of software licenses, and shortage or excessive expansion of server resources arising from existing manual-centric project operation methods. This invention can be widely applied across IT services, including development environment automation (DevOps), project management solutions, cloud auto-scaling, and software license management. Background Technology

[0004] 1. IT Project Management and Automation Technology

[0005] Existing IT project management tools (such as Jira and Trello) have primarily focused on basic functions like tracking work issues and managing schedules. In this context, dynamic resource optimization—such as workforce reallocation, adjusting software licenses, and scaling up or down server resources—is often performed manually or limited to simple threshold-based scripts. Consequently, as projects grow in scale, this can lead to bottlenecks and wasted resources.

[0006] 2. Machine Learning Prediction and Decision Support Technology

[0007] With the recent advancement of machine learning (ML) and artificial intelligence (AI) technologies, attempts are being made to predict future schedules, the volume of issues, and server load based on data. However, simple prediction models alone are insufficient because, to translate prediction results into resource allocation decisions in an actual project environment, various variables such as workforce skill mapping, software license restrictions, and cloud infrastructure cost structures must be considered.

[0008] 3. Cloud Auto Scaling Technology

[0009] The auto-scaling features provided by cloud services (AWS, Azure, GCP, etc.) offer the ability to scale server instances up or down by monitoring resource metrics such as CPU and memory usage. However, this feature primarily focuses on optimizing hardware resources and does not separately address human resource reallocation or software license management. Furthermore, it has limitations in comprehensively incorporating machine learning-based predictions, rule tables, and priority settings.

[0010] 4. Project Real-time Monitoring and Feedback Loop

[0011] A workflow that supports automated decision-making by monitoring and analyzing data such as project status (work progress, issues), server load (traffic, CPU usage), and license usage in real time, and providing immediate feedback on the results, is becoming increasingly important.

[0012] However, in existing systems, a platform capable of integrating logs and data from various sources, linking with machine learning models, and consistently performing resource control has not been sufficiently established.

[0013] 5. Problem of Inefficient Resource Allocation

[0014] ㆍ Manpower Bottleneck: Inefficiency occurs when work is concentrated on specific team members, or conversely, a large number of personnel are idle.

[0015] ㆍ Waste of SW Licenses: Cost loss due to excessive license purchase and allocation despite low actual usage.

[0016] ㆍ Hardware resource shortage or over-scaling: Server failures occur due to a surge in traffic, or conversely, costs are wasted due to excessive server expansion even when the load is low.

[0017] These problems have limitations in that they are difficult to solve solely through manual decision-making in a reality where project progress changes moment by moment.

[0018] 6. General

[0019] Considering the background technologies described above, a technical approach is required to enhance project execution efficiency by dynamically automating human, software, and hardware resources through machine learning-based prediction, rule tables, and priority setting. To overcome the limitations of these background technologies, the present invention provides an integrated solution that allocates optimal resources at each project stage and continuously corrects and improves machine learning models based on result feedback by providing an interconnected system comprising a prediction module (100), a decision engine (200), a resource control module (300), and a monitoring module (400). The problem to be solved

[0021] 1. Resolving project bottlenecks and workforce inefficiencies

[0022] In IT service development projects, bottlenecks frequently occur where work is excessively concentrated on specific team members at certain times, or conversely, idle personnel arise. This leads to project schedule delays and reduced workforce utilization. Therefore, workforce reallocation must be automated across project phases, and dynamic workforce management is required that considers technical skill sets, availability, and work overlap.

[0023] 2. Resolving SW License Over- and Under-license Issues

[0024] The simultaneous usage of various software licenses, such as build, test, and collaboration tools, is highly variable and difficult to predict during a project. A shortage of licenses causes work disruptions, while overpurchasing leads to increased unnecessary costs. Therefore, it is necessary to monitor SW license usage in real-time or periodically and dynamically increase or decrease the required quantity at the necessary time.

[0025] 3. Automation of Cloud Resource Expansion and Downscaling

[0026] Situations where server load rapidly increases or decreases at specific times due to server traffic or build pipelines are frequent. Existing threshold-based automatic scaling is limited to CPU and memory usage and fails to consider long-term demand forecasts, making it prone to over-provisioning or under-provisioning issues. There is a need for automation capabilities that combine machine learning predictions with project progress to expand or contract servers at the optimal time.

[0027] 4. Lack of alignment between forecasting and decision-making

[0028] Although predictive models using machine learning have been increasing recently, there is a problem where prediction results are not seamlessly linked to actual decision-making, such as workforce management, licensing policies, and cloud server allocation. Furthermore, while different optimizations must be pursued depending on project priorities (cost, schedule, quality, etc.), there is often a lack of systematic decision-making logic to reflect these factors.

[0029] 5. Lack of real-time monitoring and feedback

[0030] Even after resource allocation decisions are finalized, it is necessary to continuously monitor whether the allocation is actually operating efficiently and provide feedback to allow machine learning models to learn from prediction errors in order to ensure long-term accuracy improvement and project stability. However, in existing project management environments, log collection, error detection, notification systems, and error feedback are not integrated, making it difficult to fully realize the benefits of real-time optimization.

[0032] 6. Summary

[0033] To solve the above problems, the present invention aims to effectively resolve bottlenecks, inefficient resource usage, delays, and cost issues in IT service development projects by precisely estimating future demand for manpower, software, and hardware resources through a machine learning-based prediction module, calculating an optimal resource allocation plan by a decision engine reflecting rule tables and priorities, dynamically reallocating and increasing or decreasing manpower, licenses, and servers by a resource control module, and continuously improving the prediction model by receiving real-time status feedback from a monitoring module. means of solving the problem

[0035] 1. Machine learning-based prediction module (100)

[0036] Data is received from various sources, such as project management tools, server / network logs, and SW license usage history, through the data collection unit (100), and missing value processing and scaling are performed in the preprocessing unit (120).

[0037] ㆍ Using an ML model (130), future manpower demand, license demand, server load, etc. are predicted using time series analysis or regression / deep learning methods.

[0038] The feedback learning unit (140) analyzes the error between the actual measurement results and the predicted values ​​transmitted from the monitoring module (400) and the decision engine (200), and continuously corrects the model parameters using an adaptive learning or online learning method.

[0039] ㆍThe history DB (150) stores past project history and resource usage logs, which are used to train the prediction model.

[0040] 2. Decision engine (200)

[0041] • The rule table (210) defines policies necessary for resource allocation, such as personnel skill sets, server thresholds, and license limits, and the priority setting unit (220) sets priorities tailored to the project situation, such as cost minimization, schedule reduction, and quality maintenance.

[0042] The decision logic (230) combines the future resource demand forecast results of the prediction module (100), real-time monitoring data, and rule tables and priorities to finally derive resource allocation plans such as personnel reallocation, increase / decrease in SW licenses, and HW server scaling.

[0043] 3. Resource control module (300)

[0044] ㆍ As a step to actually execute the determined resource allocation and release results, the personnel management department (310) links with a project management tool to automatically assign and move the personnel to a specific task or team, and the SW license management department (320) increases or decreases the license quantity through the license server / SaaS platform API.

[0045] ㆍ The HW allocation unit (330) links with a cloud console (AWS, Azure, GCP, etc.) to increase or decrease server instances or adjust Auto Scaling settings.

[0046] ㆍ After resource control, the execution results (successful deployment, successful license purchase, successful server expansion, etc.) are reported to the monitoring module (400) and the project management tool.

[0047] 4. Monitoring module (400)

[0048] ㆍ The data collection unit (410) collects server, network, and project tool logs, etc., and the real-time analysis / notification unit (420) detects threshold exceedance or error events and promptly notifies the decision engine (200).

[0049] ㆍ The log / statistics storage unit (430) stores and analyzes long-term logs and statistical results so that they can be used to improve the accuracy of future prediction models (ML models (130)) and update rule tables (210).

[0050] This allows for constant verification of whether resource reallocation and expansion decisions are actually producing the expected effects, and enables correction of machine learning models and decision logic if errors occur.

[0051] 5. Conclusion and Operating Principle

[0052] Through the interoperability of the prediction module (100), decision engine (200), resource control module (300), and monitoring module (400) as described above, the present invention efficiently allocates and releases human, software, and hardware resources during the execution of an IT service development project, thereby resolving project bottlenecks, resource waste, and schedule delays.

[0053] Specifically, by repeatedly performing the stages of machine learning prediction, rule and priority-based decision-making, resource control, and monitoring and feedback, it is possible to respond in real-time to changes in the project environment. Furthermore, the accuracy of the prediction model improves over time through continuous learning (Adaptive / Online) techniques. This enables the simultaneous achievement of benefits such as shortened project schedules, cost reduction, and maintenance of service quality. Effects of the invention

[0054] 1. Project schedule reduction and bottleneck resolution

[0055] The present invention identifies future demand for manpower, software, and hardware resources in advance through a machine-based prediction module (100), and automatically allocates and releases necessary resources, such as manpower or servers, in a timely manner by integrating rule tables and priorities in a decision engine (200). Through this, bottlenecks caused by excessive concentration of work on specific manpower or servers are significantly reduced, and schedule delays ahead of the project are alleviated, enabling efficient progress.

[0056] 2. Cost Reduction and Maximization of Resource Utilization

[0057] Through the resource control module (300), personnel reallocation, SW license increase / decrease, and cloud server expansion / reduction are automatically performed according to the actual timing and scale required by the project. As a result, unnecessary license costs or server maintenance costs can be reduced, and idle personnel are minimized, thereby maximizing resource utilization. Consequently, there is an effect of increasing the value of project deliverables relative to cost and improving ROI (Return on Investment).

[0058] 3. Enhancement of project quality and stability

[0059] By continuously observing real-time logs and issue situations in the monitoring module (400) and providing feedback to the decision engine (200) and the prediction module (100), errors or bottlenecks can be detected early and immediate resource adjustments can be made. This allows for a rapid response to urgent issues that occur during the project, preventing a decline in service quality and increasing the overall stability of the project.

[0060] 4. Continuous performance improvement of machine learning models

[0061] The prediction module (100) periodically learns the error between actual resource usage and prediction results through adaptive learning or online learning techniques. As the model gradually improves accuracy over time, optimal resource allocation becomes possible even in long-term projects or environments with continuous high fluctuations.

[0062] 5. Reduction of management burden through integrated and automated management

[0063] Previously, project managers had to manually coordinate tasks such as workforce reallocation, SW license purchase / cancellation, and server expansion / reduction, but this invention automates and centralizes all these processes. Consequently, managers can focus more of their capabilities on strategic decision-making, and the actual workload and potential for errors are significantly reduced through API integration with various systems, such as project management tools, cloud consoles, and license servers.

[0064] 6. Scalability and Universality

[0065] This invention is not limited to specific project scales or tools and allows for various combinations of machine learning prediction techniques and rule-based decision logic, making it widely applicable from small-scale startup projects to large-scale enterprise systems. Furthermore, it can be utilized in diverse environments, such as cloud platforms (AWS, Azure, GCP, etc.), SaaS licenses, and on-premises servers, thereby universally supporting project resource optimization.

[0066] 7. Overall Effects

[0067] In conclusion, the present invention allocates and deals with manpower, software, and hardware resources at the optimal time and in an appropriate scale through machine learning prediction and dynamic automation logic, thereby achieving various effects such as shortening project schedules, reducing costs, ensuring quality and stability, reducing management burden, and enhancing model performance. This integrated resource management solution has the advantage of dramatically improving the productivity of IT service development projects and being able to flexibly respond to continuous future changes. Brief explanation of the drawing

[0069] [Drawing 1] Figure 1 schematically shows the overall configuration of the system according to the present invention. The interlinking structure between the prediction module (100), decision engine (200), resource control module (300), and monitoring module (400), and the interaction relationship with external systems such as project management tools and cloud consoles are illustrated. Through this, one can understand at a glance how human resources (people), software licenses, and hardware (servers, etc.) are integratedly automated and optimized in this invention. [Drawing 2] Figure 2 is an algorithm flowchart sequentially illustrating the algorithm flow for dynamically optimizing human resources, software, and hardware resources in the present invention. This flowchart schematically illustrates the process of dynamic resource optimization in response to changes in the project environment by repeating the steps of data collection → machine learning-based prediction (prediction module (100)) → application of rules and priorities (decision engine (200)) → resource allocation / deallocation (resource control module (300)) → monitoring / feedback (monitoring module (400)). [Drawing 3] Figure 3 illustrates a detailed view of a prediction module (100) and shows how internal components such as a data collection unit (110), a preprocessing unit (120), an ML model (130), a feedback learning unit (140), and a history DB (150) are interconnected to predict future resource requirements and improve model accuracy. You can observe the process of collecting and preprocessing data through project management tools and server / network logs, applying machine learning algorithms to calculate predicted values, and learning from errors through feedback by comparing them with actual results. [Drawing 4] Drawing 4 is a detailed drawing of a decision engine (200) and consists of a rule table (210), a priority setting unit (220), a decision logic (230), etc. Here, the future resource demand forecast received from the prediction module (100) and the real-time status provided by the monitoring module (400) are combined to finally determine the resource allocation logic regarding what personnel deployment, increase / decrease in SW licenses, and expansion / reduction of HW servers are required. [Drawing 5] Drawing 5 is a detailed drawing of a resource control module (300), illustrating the process of actual resource allocation and release by detailed functional departments such as the personnel management department (310), SW license management department (320), and HW allocation department (330) in conjunction with a project management tool, license server, cloud console, etc. The personnel management department (310) modifies the placement of specific team members, the SW license management department (320) requests an increase or decrease in licenses, and the HW allocation department (330) performs the task of automatically increasing or decreasing server instances. [Drawing 6] Drawing 6 shows a detailed view of the monitoring module (400) and is composed of a data collection unit (410), a real-time analysis / notification unit (420), a log / statistics storage unit (430), etc. Server, network, and project logs are continuously collected, and notifications can be sent to the decision engine (200) when a threshold is exceeded or an issue occurs. In addition, long-term logs and statistical information are utilized for model training of the prediction module (100) and correction of the rule table (210), thereby supporting continuous optimization of the entire system. Specific details for implementing the invention

[0070] 1. System Architecture Implementation

[0071] A. Server (Back-end) Configuration

[0072] ㆍ Project management tool integration

[0073] - Periodically retrieves task information, personnel information, and issue (bug / failure) data from collaboration tool APIs such as Jira and Trello.

[0074] - Data can be collected in real-time or periodic batches using REST APIs or Webhooks.

[0075] - The collected data is transferred to the data collection unit (110) of the prediction module (100), undergoes a preprocessing process, and is used as input data for the ML model (130).

[0076] ㆍ Cloud and License Server Integration

[0077] - Collects server instance status and scaling group configuration information by integrating with cloud console APIs such as AWS, Azure, and GCP.

[0078] - Check license usage, activation status, etc. through the SW license server or SaaS platform API.

[0079] - The information is also transmitted to the monitoring module (400) so that the real-time analysis and notification unit (420) can detect necessary events.

[0080] ㆍ Database (DB) Configuration

[0081] - Each DB or integrated DB can be operated for the history DB (150) and the log / statistics storage unit (430) (e.g., a combination of time series DB (InfluxDB, TimescaleDB) + relational DB (MySQL, etc.)).

[0082] - It allows for the long-term storage of past project logs and server load records required for machine learning model training, and enables version management by saving model update history as well.

[0084] B. Client (Front-end) Configuration

[0085] ㆍ Admin Dashboard

[0086] - Data collected and analyzed by the monitoring module (400) can be visualized to check the project progress, resource usage indicators, machine learning prediction results, etc. in real time.

[0087] - The results (resource allocation decisions) of the decision engine (200) and the execution history of the resource control module (300) may be viewed, approved, or manually adjusted.

[0088] ㆍ Notification / Event Management

[0089] - Automatically notifies administrators via various channels such as Slack, email, and SMS when thresholds are exceeded, error events, or resource shortages occur.

[0091] 2. Details of the implementation of the prediction module (100)

[0092] a. Data collection unit (110) & preprocessing unit (120)

[0093] ㆍ Data type:

[0094] - Task history, personnel allocation records, and issue occurrence data extracted from project management tools (such as Jira)

[0095] - Server monitoring logs (CPU, memory, network traffic, etc.)

[0096] - Usage history provided by the SW license server (concurrent usage, validity period, etc.)

[0097] ㆍ Preprocessing tasks:

[0098] - Remove or correct incomplete data (missing values, outliers)

[0099] - Sort in time series format based on date / time information

[0100] - Perform scaling (MinMax, Standard Scaler, etc.) and feature engineering as necessary

[0101] B. ML model (130)

[0102] ㆍ Select Model:

[0103] - Model weekly and daily resource usage patterns through time series forecasting (ARIMA, LSTM, etc.)

[0104] - Predict the relationship between project workload and manpower input based on regression (Random Forest, LightGBM, etc.)

[0105] - When complex cases (integration of human, software, and hardware resources) are required, ensemble multiple model results or apply multitasking learning.

[0106] ㆍ Online / Adaptive Learning Support:

[0107] - The feedback learning unit (140) periodically calculates the prediction error (the difference between the actual usage and the predicted value).

[0108] - If the error exceeds a certain threshold, rebalance Hyperparameters or Weights (Adaptive Learning)

[0109] - If there is a large amount of real-time streaming data, you can also use online learning frameworks (e.g., Vowpal Wabbit, Flink ML, etc.).

[0110] ㆍ History DB (150)

[0111] - By long-term storage of past project history, model training data, and feedback logs, it contributes to improving initial prediction accuracy by utilizing existing patterns when resuming projects or launching new ones.

[0113] 3. Details of the implementation of the decision engine (200)

[0114] a. Rule table (210)

[0115] Manpower Skill Set Mapping:

[0116] - Examples: "Frontend Development: Team Members A, B", "DB Tuning: Team Member C", etc.

[0117] - You can also specify whether multiple people can be deployed simultaneously based on priority (cost, schedule), vacation schedules, etc.

[0118] ㆍ Server threshold:

[0119] - Basic rules such as "Add one server if CPU usage exceeds 80% for more than 5 minutes"

[0120] - Also used as a reference when the decision engine makes additional judgments by correcting machine learning prediction results

[0121] ㆍ License Limit:

[0122] - "CI / CD license allows simultaneous use by up to 10 people," etc.

[0123] b. Priority setting unit (220)

[0124] ㆍ Set whether to weight cost, schedule, or quality, or to apply composite weights.

[0125] ㆍ Weights may change dynamically depending on the project phase (initial, mid, or late).

[0126] C. Decision logic (230)

[0127] ㆍReceive the result (resource demand at a specific future point in time) of the prediction module (100) as input and combine it with the rule table (210) and priority (220).

[0128] ㆍ Conditional determination of workforce reallocation, increases / decreases in SW licenses, expansion / reduction of HW servers, etc.

[0129] ㆍ If necessary, go through the administrator approval process, or if the threshold is not high, instruct the resource control module (300) directly through automatic approval.

[0130] 4. Details of resource control module (300) implementation

[0131] a. Human Resources Management Department (310)

[0132] ㆍ Connect with project management tool APIs to assign or release tasks to specific team members, or move personnel between teams

[0133] ㆍ When work reallocation is performed, the corresponding event (e.g., "Personnel X → Additional deployment to Team A") is recorded in the monitoring module (400) and the administrator dashboard.

[0134] b. SW License Management Department (320)

[0135] ㆍ Expand or reduce the number of concurrent users using the License Server (SaaS platform) API

[0136] You can configure payment / billing integration to be processed automatically or to proceed after administrator approval.

[0137] ㆍ Reduce costs by releasing some licenses if there is an overload.

[0138] c. HW allocation unit (330)

[0139] Call server instance creation / deletion commands in the cloud console (AWS, Azure, GCP, etc.)

[0140] ㆍ Change load balancer (ELB, ALB, etc.) or Auto Scaling group settings if necessary

[0141] Monitor whether the actual service has been smoothly updated after server expansion, and if any issues arise, immediately roll back or retry.

[0142] 5. Details of the implementation of the monitoring module (400)

[0143] a. Data collection unit (410)

[0144] ㆍ Integrated collection of server / network logs (CPU, memory, disk, traffic), project management tool issue / task logs, license usage logs, etc.

[0145] You can use streaming methods such as Kafka, MQTT, and Webhooks, or schedule them at fixed intervals.

[0146] B. Real-time analysis / notification unit (420)

[0147] ㆍ Immediately identify events such as exceeding a threshold, failure, or work delay, and notify the decision engine (200).

[0148] Notifications can be sent to administrators via Slack, email, SMS, etc.

[0149] c. Log / statistics storage unit (430)

[0150] ㆍ Store long-term logs, resource control history, usage statistics, etc. in a DB or dashboard

[0151] ㆍ It is also used for feedback learning of the prediction module (100) and can be used as retrospective analysis data after the project ends.

[0152] 6. Examples of Actual Application Scenarios

[0153] A. Mid-stage of development

[0154] ㆍ The ML model (130) predicts "a 20% increase in frontend workload over the next 2 weeks" → The decision engine (200) decides to add one person with frontend skills to Team A → The personnel management department (310) of the resource control module (300) performs the reallocation via the Jira API.

[0155] B. Testing Phase

[0156] ㆍ It is determined that the server load is higher than predicted → The decision engine (200) decides to increase the number of HW servers by 2 → The HW allocation unit (330) of the resource control module (300) issues an increase command to the AWS Auto Scaling group → The monitoring module (400) checks whether the new servers are operating normally and feeds back the actual load data to the prediction module (100).

[0157] C. Short-term expansion of SW licenses

[0158] ㆍ Overload occurred as concurrent usage of build / deployment tools was 2 users higher than predicted → Immediately increased licenses by 2 → Automatically released when usage decreases after a certain period.

[0159] 7. Effects and Benefits

[0160] A. Improvement of project progress efficiency: Predicting and responding in advance to manpower bottlenecks, hardware traffic overload, software license shortages, etc.

[0161] B. Cost Reduction: Reduces over-provisioning issues as idle personnel, licenses, and servers are minimized.

[0162] C. Continuous Model Improvement: Increased long-term prediction accuracy by training machine learning models online with monitoring feedback

[0163] D. Automation and Real-time Response: Fully automated from decision-making to execution and monitoring, significantly reducing the workload of project managers.

[0165] 8. Conclusion

[0166] As described above, by linking the prediction module (100), decision engine (200), resource control module (300), and monitoring module (400) together and implementing dynamic resource optimization through machine learning-based prediction, human, software, and hardware resources in an IT service development project can be optimized in real time and periodically.

[0167] The present invention is flexibly applicable to project scale and environment (cloud / on-premises, SaaS / internal license, etc.), simultaneously achieving project schedule reduction and cost reduction, as well as significantly improving system reliability and quality in the long term.

[0168] Explanation of the symbols

[0169] 1. (100) Prediction Module ㆍ (110) Data collection unit - Periodically or in real-time acquire raw data from various sources, such as project management tools, server / network logs, and software license records. ㆍ (120) Preprocessing section - Transform collected raw data into a format suitable for machine learning, such as handling missing and outlier values, scaling, and feature engineering. ㆍ (130) ML Model - Predict future usage of manpower, software, and hardware resources through machine learning techniques such as time series analysis, regression, and deep learning ㆍ (140) Feedback Learning Department - Analyze the error between the actual data and the predicted value transmitted from the monitoring module (400) and the decision engine (200) to continuously improve model performance through online learning or adaptive learning. ㆍ (150) History DB - Stores past project history, resource usage logs, training data, etc., to utilize for model training and future prediction refinement. 2. (200) Decision Engine ㆍ (210) Rule Table - Stores policy and rule information necessary for resource allocation, such as workforce skill sets, server thresholds, and license restrictions ㆍ (220) Priority setting section ​- Dynamically assign goal priorities such as "cost minimization," "schedule adherence," and "quality maintenance" based on project conditions ㆍ (230) Decision Logic - The prediction results of the prediction module (100), the rule table (210), and the priority setting unit (220) are combined to finally determine resource allocation plans such as personnel placement, increase / decrease in SW licenses, and expansion / reduction of HW servers. 3. (300) Resource Control Module ㆍ (310) Human Resources Department - Automate the reallocation of work for specific personnel and the assignment and release of tasks through project management tool (such as Jira) APIs. ㆍ (320) SW License Management Department - Manage license expansion / reduction and usage status through integration with license server or SaaS platform APIs. ㆍ (330) HW Allocation Section - Perform hardware resource allocation and deallocation, such as creating and deleting server instances and changing Auto Scaling group settings, in conjunction with cloud consoles (AWS, Azure, GCP, etc.). 4. (400) Monitoring Module ㆍ (410) Data Collection Department - Collect operational data and issue information, such as server / network / project tool logs and resource control results, in real-time and periodically. ㆍ (420) Real-time Analysis / Notification Department - Detecting threshold exceedance, failure occurrence, delay situation, etc., and immediately sending a notification to the decision engine (200) or notifying the administrator via Slack / email / SMS, etc. ㆍ (430) Log / Statistics Storage Unit - Stores and analyzes long-term log and event history and statistical information to be used for model training of the prediction module (100) and project performance measurement.

Claims

Claim 1 A method for dynamically optimizing human, software, and hardware resources in an IT service development project, comprising: a step of collecting personnel information, software license usage information, and hardware resource status through a project management tool and server / network logs, etc.; a step in which a machine learning-based prediction module (100) predicts the personnel, license, and server resources required at a specific point in the future based on the collected data; a step in which a decision engine (200) including the prediction results, a rule table (210), and a priority setting unit (220) determines whether to reallocate personnel, increase or decrease licenses, or expand or reduce servers; a step in which a resource control module (300) actually allocates or releases the decisions of the decision engine (200) through a project management tool API, a cloud console, etc.; and a step in which a monitoring module (400) collects and analyzes the actual usage results after the resource allocation / release to identify the prediction error and feedback-learn the machine learning model, thereby automatically and dynamically optimizing human, software, and hardware resources according to the project progress stage. Claim 2 A method according to claim 1, wherein the prediction module (100) performed in step (2) comprises: a. acquiring raw data from a project management tool, a server monitoring system, a license server, etc., from a data collection unit (110); b. performing preliminary operations such as processing missing values, removing outliers, and scaling through a preprocessing unit (120); c. calculating future resource requirements by applying time series analysis or regression / deep learning techniques to an ML model (130); and d. a feedback learning unit (140) analyzing the error between actual usage and the predicted value and readjusting model parameters through adaptive learning or online learning. Claim 3 A method according to claim 1 or claim 2, wherein the decision engine (200) performed in step (3) refers together with a rule table (210) containing personnel skill sets and schedule information, and a priority setting unit (220) based on the project status to determine which goal to prioritize among cost minimization, schedule compliance, and quality maintenance, and calculates resource allocation logic by combining with the prediction result. Claim 4 A method according to any one of claims 1 to 3, wherein the resource control module (300) performed in step (4) is characterized by: ㆍ (i) changing the task assignment of a specific personnel or performing reassignment between teams via the API of a project management tool through the personnel management unit (310); ㆍ (ii) increasing or decreasing the license quantity via the API of a license server or SaaS platform through the SW license management unit (320); and ㆍ (iii) automatically allocating and releasing personnel, software, and hardware resources by controlling a cloud console (AWS, Azure, GCP, etc.) through the HW allocation unit (330) to increase or decrease server instances or change Auto Scaling settings. Claim 5 A method according to any one of claims 1 to 4, wherein the monitoring module (400) performed in step (5) comprises: (a) collecting server, network, project tool logs and resource control history in real time from a data collection unit (410); (b) detecting threshold exceedance or error events, etc., from a real-time analysis / notification unit (420) and immediately notifying the decision engine (200); and (c) accumulating long-term logs and statistics through a log / statistics storage unit (430) and utilizing them for model updates of the prediction module (100) and improvement of the rule table (210). Claim 6 A dynamic resource optimization system configured to automatically optimize human, software, and hardware resources by integrating the following: (A) a prediction module (100): a means for predicting future resource demand using machine learning based on past and present project data, server and license usage logs, etc., and retraining the model through error feedback; (B) a decision engine (200): a means for determining whether to allocate resources by integrating the prediction results, rule tables, and priorities provided in (A); (C) a resource control module (300): a means for actually increasing, decreasing, or reallocating human, software, and hardware resources by linking with project management tools, license servers, cloud consoles, etc., according to the decisions of (B); and (D) a monitoring module (400): a means for collecting and analyzing results performed in (C) and server, network, and project progress logs in real time, and providing feedback and notifications to (A) and (B); wherein (A) to (D) collaborate to automatically optimize human, software, and hardware resources in an IT service development project. Claim 7 A computer-readable recording medium storing a program for dynamically optimizing human, software, and hardware resources, characterized in that a computer executing the program is configured to perform the following:

1. a step of collecting and preprocessing information on project management tools, server and network logs, and license usage; 2. a step of predicting future resource requirements through a machine learning model and calculating resource allocation logic by referencing rules and priorities; 3. a step of increasing, decreasing, or reallocating human, software, and hardware resources through a resource control module; 4. a step of collecting and analyzing actual usage results in a monitoring module, receiving feedback on prediction errors, and continuously updating the machine learning model; and 5. a recording medium that records a program for automatically and dynamically optimizing human resources, software licenses, and hardware servers in an IT service development project by repeating steps 1) through 4) periodically or in real time.

Images