Project risk dynamic management method, system and equipment based on big data

Abstract

The invention discloses a project risk dynamic management method, system and equipment based on big data, and the method comprises the steps: obtaining all material data generated in a project whole-process management process based on a constructed project management system, carrying out the data processing of original data through an information processing module, and constructing a project risk database; performing project risk prediction by applying a long short-term memory (LSTM) network based on the project risk database, and judging whether the project has a risk at the moment and a corresponding risk level; if the project has the risk, starting a project risk type identification algorithm, performing risk category identification based on project data by applying a gradient boosting decision tree (GBDT) algorithm, and determining a project risk category; and based on the risk level and type of the project, matching a project risk database to determine risk countermeasures, and pushing the risk countermeasures to related managers in real time. According to the method, the prediction model and the identification model are combined, so that the risk category can be identified more accurately, and the prediction and identification rate of the project risk is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet of Things project management, and in particular to a project risk dynamic management method, system and equipment based on big data. Background Art

[0002] With the interconnection and integration of technologies such as the Internet of Things, cloud computing, and artificial intelligence, the development and management of highly integrated and intelligent Internet of Things projects have become increasingly complex. In the process of project development and management, there are many uncertainties such as technology, internal management, and external environment. As a result, projects face various risks during the process of advancement, such as cost overruns, schedule delays, mismatch between requirements and functions, and non-cooperation from suppliers. If risks are not discovered and eliminated in a timely manner, project delays or even failures will occur.

[0003] At present, the risk management of IoT projects mainly relies on the experience of project managers to evaluate the possibility of related risks and the degree of risk impact, and then determine the risk level through statistical analysis. This qualitative analysis method is easy to operate, but because it is based on subjective judgment, it is highly subjective, the risk management level is uneven, and it has a certain degree of arbitrariness. In addition, there are a large number of uncertain factors in the management process of the development and implementation of IoT projects. Risks show obvious dynamic changes over time, and the generation and action process of risks are likely to have the characteristics of "fluidity". At present, traditional project risk management methods can only identify and manage static risks, and cannot handle dynamic risks caused by uncertainty in the time dimension. In addition, they mainly rely on the experience of managers for qualitative identification and judgment, and the accuracy of risk identification is low and the timeliness is poor.

[0004] Therefore, building a quantitative and dynamic project risk management method and system based on big data to realize the automated dynamic identification of project risks will help to identify project risks in a timely and accurate manner, improve project management level, and reduce project management costs.

[0005] Related terms explanation

[0006] LSTM (Long Short-Term Memory): A time recurrent neural network designed to solve the long-term dependency problem of general RNNs (recurrent neural networks). All RNNs have a chain form of repeated neural network modules.

[0007] GBDT (Gradient Boosting Decision Tree): A decision tree-based ensemble learning algorithm that builds a set of weak learners (decision trees) and accumulates their prediction results as the final prediction output. This algorithm combines the ideas of gradient boosting and decision trees to gradually improve prediction accuracy through iteration.

[0008] CART regression tree (Classification and Regression Tree): A decision tree learning algorithm that can be used to build both classification and regression trees. The CART regression tree is a decision tree generated by the CART algorithm when the classification to be predicted is continuous data. Summary of the invention

[0009] The present invention proposes a project risk dynamic management method, system and device based on big data, which can solve at least one of the technical problems in the background technology.

[0010] To achieve the above object, the present invention adopts the following technical solutions:

[0011] A project risk dynamic management method based on big data, comprising the following steps:

[0012] Based on the constructed project management system, all material data generated during the whole process of project management are obtained, the original data is processed through the information processing module, and a project risk database is constructed;

[0013] Based on the project risk database, the long short-term memory network LSTM is used to predict project risks and determine whether the project has risks at this moment and the corresponding risk level;

[0014] If there are risks in the project, start the project risk type identification algorithm, apply the gradient boosting decision tree (GBDT) algorithm based on the project data to identify the risk category and determine the project risk category;

[0015] Based on the risk level and type of the project, the risk response measures are determined by matching the project risk database and pushed to relevant managers in real time.

[0016] Furthermore, the construction of the project risk database includes firstly collecting Chinese and English literature data in the field of project risk based on Python crawler technology, applying text big data analysis methods and project management theory, and constructing a set of characteristic factors for project risk prediction, including technical characteristic factors, demand characteristic factors, personnel characteristic factors, schedule characteristic factors, cost characteristic factors, contract characteristic factors, financial characteristic factors, collection characteristic factors and procurement characteristic factors;

[0017] Then, a project risk database is constructed based on the characteristic factor set, covering dimensional data such as project technical data, demand data, personnel data, progress data, cost data, contract data, financial data, collection data and procurement data, and time series data is established based on the time dimension information of regularly updated data.

[0018] Furthermore, based on the project risk database, the long short-term memory network LSTM is used to predict project risks. First, a risk dynamic prediction model is constructed based on the project risk database and the long short-term memory network LSTM algorithm, including:

[0019] Model training data: Model training is performed based on the completed project data in the project risk database. The sequence of completed projects input at time t is:

[0020]

[0021] in represents the risk factor parameter of project p at time t, i = 1, 2, ..., 9; p = 1, 2, ..., P; the data of completed projects are dynamically updated in real time with the project delivery progress;

[0022] Apply the long short-term memory network LSTM algorithm to X(t) for model training. The steps of the LSTM algorithm are as follows:

[0023] Forget gate: f t Indicates the state of the forget gate at time t, controls the degree of forgetting of the memory at the previous moment, and determines which information should be abandoned or retained;

[0024] f t =σ(W f ·[h t-1 ,x(t)]+b f )

[0025] Where x(t) represents the input of the unit at time t, h t-1 represents the output of the unit at time t-1, b f Represents the bias term of the forget gate, W f represents the weight matrix of the forget gate, σ represents the sigmoid neuron function, and the function sigmoid is defined as

[0026]

[0027] Input gate: i t Represents the input gate, which determines the information update status required by the unit at the current moment. Indicates the update status of backup information, C tIndicates the unit status at the current moment, controls the updating of important information and the discarding of useless information;

[0028]

[0029] The tanh function is defined as

[0030]

[0031] Output gate: o t Represents the state of the output gate, which is used to determine the value of the next hidden state and controls the influence of long-term memory on the current output;

[0032] h t =o t ×tanh(C t )

[0033] o t =σ(W o ·[h t-1 ,x(t)]+b o )

[0034] Risk prediction of ongoing projects: Input the series data of project p at time t

[0035]

[0036] X p (t)′ is input into the LSTM algorithm for risk prediction, and the project risk value h′ is output in real time t .

[0037] Furthermore, it is determined whether the project has risks at this moment and the corresponding risk levels include:

[0038] Based on the completed project risk index data in the project risk database, determine the mean risk of completed projects at time t As the risk threshold, When , it is considered that the ongoing project p is at risk.

[0039] Furthermore, if there are risks in the project, the project risk type identification algorithm is started, and the risk category is identified by applying the gradient boosting decision tree GBDT algorithm based on the project data. Determining the project risk category includes applying the gradient boosting decision tree GBDT algorithm to perform model training based on the completed project data in the project risk database. The steps are as follows:

[0040] (1) Initialize the first weak learner F0(x):

[0041]

[0042] Among them, P(Y=1|x) is the proportion of y=1 in the training samples, and the learner is initialized using prior information;

[0043] (2) Establish M classification and regression trees m = 1, 2, ..., M;

[0044] For i=1,2,…,N, calculate the response value corresponding to the mth tree, the negative gradient of the loss function, that is, the pseudo residual:

[0045]

[0046] For i = 1, 2, ..., N, use CART regression tree to fit the data (x i ,r m,i ), we get the mth regression tree, whose corresponding leaf node area is R m,j , where j = 1, 2, ..., J m , and J m is the number of child nodes of the th regression tree;

[0047] For J m leaf node regions j=1,2,…,J m , calculate the best fit value:

[0048]

[0049] Update the strong learner F m (x):

[0050]

[0051] The final strong learner F M (x) is expressed as:

[0052]

[0053] The gradient boosting decision tree GBDT classification algorithm uses a series of gradient boosting trees to fit the logarithmic probability, and finally outputs the probability of comparing categories. The larger the probability, the more likely it is to be predicted as that category. Its model is expressed as:

[0054]

[0055] When it is predicted that there are risks in the ongoing project p, the data of the ongoing project p is input into the trained gradient boosting decision tree algorithm for automatic identification of risk categories. The corresponding risk categories are technical risk, demand risk, personnel risk, schedule risk, cost risk, contract risk, financial risk, collection risk and procurement risk.

[0056] On the other hand, the present invention also discloses a project risk dynamic management system based on big data, which is used to implement the above-mentioned project risk dynamic management method based on big data, including a project management system, a data processing module, a project risk database module, a risk dynamic prediction module, a risk category automatic identification module, and a risk early warning management module;

[0057] The project management system is used to obtain all material data generated during the whole process of project management;

[0058] The data processing module is used to process the original data through the information processing module;

[0059] The project risk database module is used to build a project risk database;

[0060] The risk dynamic prediction module is used to predict project risks based on the project risk database using the long short-term memory network LSTM, and to determine whether the project has risks at this moment and the corresponding risk level;

[0061] The risk category automatic identification module is used to start the project risk type identification algorithm if there is a risk in the project, and apply the gradient boosting decision tree GBDT algorithm based on the project data to identify the risk category and determine the project risk category;

[0062] The risk warning management module determines risk response measures based on the risk level and type of the project, matches the project risk database, and pushes them to relevant managers in real time.

[0063] On the other hand, the present invention further discloses a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the above method.

[0064] On the other hand, the present invention further discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the above method.

[0065] It can be seen from the above technical scheme that the present invention discloses a method, system and equipment for dynamic management of project risks based on big data, which are mainly implemented through the following steps: first, based on the constructed project management system, all material data generated in the whole process of project management are obtained, the original data is processed through the information processing module, and a project risk database is constructed; second, based on the project risk database, a long short-term memory network (LSTM) is used to predict project risks, and it is determined whether the project has risks at this moment and the corresponding risk level; third, if there are risks in the project, the project risk type identification algorithm is started, and the gradient boosting decision tree (GBDT) algorithm is applied based on the project data to identify the risk category and determine the project risk category; fourth, based on the risk level and type of the project, the project risk database is matched to determine the risk response measures, and they are pushed to relevant managers in real time.

[0066] The present invention creatively proposes to build a set of continuously dynamically updated project risk database based on historical project data, forming an intelligent risk quantification management framework driven by big data, making project risk management more empirical, data models more scientific, and data standards more standardized, improving the scientific nature of project risk analysis methods and data decision-making models, and enriching the knowledge structure, data structure, and management model of project risk management. And for the first time, the long short-term memory network (LSTM) and gradient boosting decision tree (GBDT) algorithm are used to build an integrated risk prediction and identification model to perform dynamic prediction and type identification of project risks. The model has the ability to automatically hierarchically learn features and analyze the deep relationship between information, can handle complex nonlinear problems, can fully explore the inherent fluctuation information and long-term dependencies between project data, improve the prediction and identification rate of project risks, and make the process of project risk management more dynamic and accurate.

[0067] Specifically, the beneficial effects of the present invention are as follows:

[0068] 1. This invention uses long short-term memory network and gradient boosting decision tree algorithm to construct an integrated risk prediction and identification model for the first time, and performs dynamic prediction and type identification of project risks. The model can handle complex nonlinear problems and fully explore the deep long-term dependencies between project data information. Compared with traditional algorithms, it has higher generalization ability and prediction and identification accuracy. Compared with the existing technical solutions, combining the prediction model and the identification model can more accurately identify risk categories and improve the prediction and identification rate of project risks.

[0069] 2. The present invention constructs a continuously and dynamically updated project risk database based on historical project data, forming an intelligent risk quantification management framework driven by big data. It can perform quantitative risk analysis based on project data and identify unpredictable risks, solving the inaccurate traditional method of relying on the experience of managers for qualitative identification and judgment, improving the scientific nature of project risk analysis methods and data decision-making models, and enriching the knowledge structure, data structure and management model of project risk management.

[0070] 3. The present invention integrates the project management system, project risk database, project risk prediction and identification algorithm into one platform to realize dynamic data update and automatic risk identification, and can match corresponding solutions according to risk types to form an integrated project risk management system, thus realizing the transformation of project risks from manual identification to automatic identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 is a system block diagram of the present invention;

[0072] Figure 2 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0073] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments.

[0074] like Figure 1 As shown, the project risk dynamic management system based on big data of the present invention includes a project management system, a data processing module, a project risk database module, a risk dynamic prediction module, a risk category automatic identification module, and a risk early warning management module;

[0075] The project management system is used to obtain all material data generated during the whole process of project management;

[0076] The data processing module is used to process the original data through the information processing module;

[0077] The project risk database module is used to build a project risk database;

[0078] The risk dynamic prediction module is used to predict project risks based on the project risk database using the long short-term memory network LSTM, and to determine whether the project has risks at this moment and the corresponding risk level;

[0079] The risk category automatic identification module is used to start the project risk type identification algorithm if there is a risk in the project, and apply the gradient boosting decision tree GBDT algorithm based on the project data to identify the risk category and determine the project risk category;

[0080] The risk warning management module determines risk response measures based on the risk level and type of the project, matches the project risk database, and pushes them to relevant managers in real time.

[0081] The relationship between the modules of the project risk dynamic management system based on big data and the dynamic management methods corresponding to the application scenarios are as follows: Figure 2 As shown:

[0082] Step 1: Build a project management system for the full process management of the project, including the business opportunity stage, solution stage, bidding stage, startup stage, planning stage, implementation stage, closing stage and maintenance stage. The system covers all materials in the project management process, including technical documents, contract information, approval information, time information, personnel information, cost information, financial information, collection information, procurement information and other data. Project managers continuously input relevant data of each project into the project management system and update it dynamically in real time. At the same time, collect risk treatment solutions from various product teams and departments, and build a project risk response solution library based on historical project risk treatment solutions, and update it in real time according to changes in product capabilities, management methods and other factors.

[0083] Step 2: The data processing module analyzes, extracts, and cleans various structures and types of data in the project management system according to the requirements of the project risk database, extracts valid data, and builds a multi-dimensional data parameter database.

[0084] Step 3: Construct a project risk database. First, collect Chinese and English literature data in the field of project risk based on Python crawler technology, apply text big data analysis methods and project management theory, and construct a set of feature factor sets for project risk prediction, including technical feature factors, demand feature factors, personnel feature factors, schedule feature factors, cost feature factors, contract feature factors, financial feature factors, collection feature factors, and procurement feature factors. Then, build a project risk database based on the feature factor set, covering project technical data, demand data, personnel data, schedule data, cost data, contract data, financial data, collection data, and procurement data, and establish time series data based on the time dimension information of the data being updated regularly.

[0085] Step 4: Build a risk dynamic prediction model based on the project risk database and LSTM algorithm:

[0086] Model training data: Model training is performed based on the completed project data in the project risk database. The sequence of completed projects input at time t is:

[0087]

[0088] in It represents the risk factor parameter of project p at time t, i = 1, 2, ..., 9; p = 1, 2, ..., P. The data of completed projects are updated in real time with the delivery progress of the project.

[0089] Apply LSTM algorithm to X(t) for model training. The working principle of LSTM algorithm mainly consists of the following update process:

[0090] Forget gate: f t Represents the state of the forget gate at time t, controls the degree of forgetting of the memory at the previous moment, and determines which information should be abandoned or retained.

[0091] f t =σ(W f ·[h t-1 ,x(t)]+b f )

[0092] Where x(t) represents the input of the unit at time t, h t-1 represents the output of the unit at time t-1, b f Represents the bias term of the forget gate, W f represents the weight matrix of the forget gate, σ represents the sigmoid neuron function, and the function sigmoid is defined as

[0093]

[0094] Input gate: i t Represents the input gate, which determines the information update status required by the unit at the current moment. Indicates the update status of backup information, C t Indicates the unit status at the current moment, controls the updating of important information and the discarding of useless information.

[0095]

[0096] The tanh function is defined as

[0097]

[0098] Output gate: o t Represents the state of the output gate, which is used to determine the value of the next hidden state and controls the influence of long-term memory on the current output.

[0099] h t =ot ×tanh(C t )

[0100] o t =σ(W o ·[h t-1 ,x(t)]+b o )

[0101] Risk prediction of ongoing projects: Input the series data of project p at time t

[0102]

[0103] x p (t)′ is input into the LSTM algorithm for risk prediction, and the project risk value h′ is output in real time t .

[0104] Project risk assessment: Based on the completed project risk index data in the project risk database, determine the average risk of completed projects at time t. As the risk threshold, When , it is considered that the ongoing project p is at risk.

[0105] Step 5: Automated risk category identification module. The present invention applies the gradient boosting decision tree (GBDT) algorithm to perform model training based on the completed project data in the project risk database. The algorithm can flexibly process various types of data, has high prediction accuracy and good generalization performance.

[0106] (1) Initialize the first weak learner F0(x):

[0107]

[0108] Among them, P(Y=1|x) is the proportion of y=1 in the training samples, and the prior information is used to initialize the learner.

[0109] (2) Establish M classification and regression trees m = 1, 2, ..., M;

[0110] For i=1,2,…,N, calculate the response value corresponding to the mth tree (the negative gradient of the loss function, i.e., the pseudo residual):

[0111]

[0112] For i = 1, 2, ..., N, use CART regression tree to fit the data (x i ,r m,i ), we get the mth regression tree, whose corresponding leaf node area is R m,j , where j = 1, 2, ..., J m , and J mis the number of child nodes of the th regression tree.

[0113] For J m leaf node regions j=1,2,…,J m , calculate the best fit value:

[0114]

[0115] Update the strong learner F m (x):

[0116]

[0117] The final strong learner F M (x) can be expressed as:

[0118]

[0119] The GBDT classification algorithm uses a series of gradient boosting trees to fit the logarithmic probability, and finally outputs the probability of comparing categories. If the probability is large, it is predicted to be that category. Its model can be expressed as:

[0120]

[0121] When it is predicted that there are risks in the ongoing project p, the data of the ongoing project p is input into the trained gradient boosting decision tree algorithm for automatic identification of risk categories. The corresponding risk categories are technical risk, demand risk, personnel risk, schedule risk, cost risk, contract risk, financial risk, collection risk and procurement risk.

[0122] Step 6: Risk warning module, based on the risk category automatic identification model, identifies the project risk type and uploads the identification results to the monitoring platform for output display. If the identification result is abnormal, a warning signal will be issued and the information will be pushed to the project manager for confirmation; if the risk is confirmed, the corresponding response plan will be matched based on the risk response plan library of the platform project management system.

[0123] In summary, the present invention proposes to construct a set of continuously dynamically updated project risk database based on project historical data, forming an intelligent risk quantification management framework driven by big data, making project risk management more empirical, data models more scientific, and data standards more standardized, improving the scientific nature of project risk analysis methods and data decision-making models, and enriching the knowledge structure, data structure, and management model of project risk management. And for the first time, the long short-term memory network (LSTM) and gradient boosting decision tree (GBDT) algorithm are used to construct an integrated risk prediction and identification model to perform dynamic prediction and type identification of project risks. The model has the ability to automatically hierarchically learn features and analyze the deep relationship between information, can handle complex nonlinear problems, can fully explore the inherent fluctuation information and long-term dependencies between project data, improve the prediction and identification rate of project risks, and make the process of project risk management more dynamic and accurate.

[0124] On the other hand, the present invention further discloses a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the steps of the above method.

[0125] On the other hand, the present invention further discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the above method.

[0126] In another embodiment provided in the present application, a computer program product including instructions is also provided, which, when executed on a computer, enables the computer to execute any of the big data-based dynamic project risk management methods in the above-mentioned embodiments.

[0127] It is understandable that the system, device and storage medium provided in the embodiments of the present invention correspond to the method provided in the embodiments of the present invention, and the explanation, examples and beneficial effects of the relevant contents can refer to the corresponding parts in the above methods.

[0128] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium may be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive Solid State Disk (SSD)), etc.

[0129] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0130] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0131] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A project risk dynamic management method based on big data, characterized in that: The following steps are included: Based on the constructed project management system, all material data generated during the whole process of project management are obtained, the original data is processed through the information processing module, and a project risk database is constructed; Based on the project risk database, the long short-term memory network LSTM is used to predict project risks and determine whether the project has risks at this moment and the corresponding risk level; If there are risks in the project, start the project risk type identification algorithm, apply the gradient boosting decision tree (GBDT) algorithm based on the project data to identify the risk category and determine the project risk category; Based on the risk level and type of the project, the risk response measures are determined by matching the project risk database and pushed to relevant managers in real time.

2. The project risk dynamic management method based on big data according to claim 1 is characterized by: The construction of the project risk database includes firstly collecting Chinese and English literature data in the field of project risk based on Python crawler technology, applying text big data analysis methods and project management theory, and constructing a set of characteristic factors for project risk prediction, including technical characteristic factors, demand characteristic factors, personnel characteristic factors, schedule characteristic factors, cost characteristic factors, contract characteristic factors, financial characteristic factors, collection characteristic factors and procurement characteristic factors; Then, a project risk database is constructed based on the characteristic factor set, covering dimensional data such as project technical data, demand data, personnel data, progress data, cost data, contract data, financial data, collection data and procurement data, and time series data is established based on the time dimension information of regularly updated data.

3. The project risk dynamic management method based on big data according to claim 1 is characterized by: Based on the project risk database, the long short-term memory network LSTM is used to predict project risks. First, a risk dynamic prediction model is built based on the project risk database and the long short-term memory network LSTM algorithm, including: Model training data: Model training is performed based on the completed project data in the project risk database. The sequence of completed projects input at time t is: in represents the risk factor parameter of project p at time t, i = 1, 2, ..., 9; p = 1, 2, ..., P; the data of completed projects are dynamically updated in real time with the project delivery progress; Apply the long short-term memory network LSTM algorithm to X(t) for model training. The steps of the LSTM algorithm are as follows: Forget gate: f t Indicates the state of the forget gate at time t, controls the degree of forgetting of the memory at the previous moment, and determines which information should be abandoned or retained; f t =σ(E f ·[h t-1 ,x(t)]+b f ) Where x(t) represents the input of the unit at time t, h t-1 represents the output of the unit at time t-1, b f Represents the bias term of the forget gate, W f represents the weight matrix of the forget gate, σ represents the sigmoid neuron function, and the function sigmoid is defined as Input gate: i t Represents the input gate, which determines the information update status required by the unit at the current moment. Indicates the update status of backup information, C t Indicates the unit status at the current moment, controls the updating of important information and the discarding of useless information; The tanh function is defined as Output gate: o t Represents the state of the output gate, which is used to determine the value of the next hidden state and controls the influence of long-term memory on the current output; h t =o t ×tanh(C t ) o t =σ(W o ·[h t-1 ,x(t)]+b o ) Risk prediction of ongoing projects: Input the series data of project p at time t X p (t)′ is input into the LSTM algorithm for risk prediction, and the project risk value h′ is output in real time t .

4. The project risk dynamic management method based on big data according to claim 3 is characterized by: And determine whether the project has risks at this moment and the corresponding risk levels include: Based on the completed project risk index data in the project risk database, determine the mean risk of completed projects at time t As the risk threshold, When , it is considered that the ongoing project p is at risk.

5. The project risk dynamic management method based on big data according to claim 1 is characterized by: If there is a risk in the project, start the project risk type identification algorithm, and apply the gradient boosting decision tree GBDT algorithm based on the project data to identify the risk category. Determining the project risk category includes applying the gradient boosting decision tree GBDT algorithm to the completed project data in the project risk database for model training. The steps are as follows: (1) Initialize the first weak learner F0(x): Among them, P(Y=1|x) is the proportion of y=1 in the training samples, and the learner is initialized using prior information; (2) Establish M classification and regression trees m = 1, 2, ..., M; For i=1,2,…,N, calculate the response value corresponding to the mth tree, the negative gradient of the loss function, that is, the pseudo residual: For i = 1, 2, ..., N, use CART regression tree to fit the data (x i ,r m,i ), we get the mth regression tree, whose corresponding leaf node area is R m,j , where j = 1, 2, ..., J m , and J m is the number of child nodes of the th regression tree; For J m leaf node regions j=1,2,…,J m , calculate the best fit value: Update the strong learner F m (x): The final strong learner F M (x) is expressed as: The gradient boosting decision tree GBDT classification algorithm uses a series of gradient boosting trees to fit the logarithmic probability, and finally outputs the probability of comparing categories. The larger the probability, the more likely it is to be predicted as that category. Its model is expressed as: When it is predicted that there are risks in the ongoing project p, the data of the ongoing project p is input into the trained gradient boosting decision tree algorithm for automatic identification of risk categories. The corresponding risk categories are technical risk, demand risk, personnel risk, schedule risk, cost risk, contract risk, financial risk, collection risk and procurement risk.

6. A project risk dynamic management system based on big data, used to implement the project risk dynamic management method based on big data according to any one of claims 1 to 5, characterized in that: It includes project management system, data processing module, project risk database module, risk dynamic prediction module, risk category automatic identification module and risk early warning management module; The project management system is used to obtain all material data generated during the whole process of project management; The data processing module is used to process the original data through the information processing module; The project risk database module is used to build a project risk database; The risk dynamic prediction module is used to predict project risks based on the project risk database using the long short-term memory network LSTM, and to determine whether the project has risks at this moment and the corresponding risk level; The risk category automatic identification module is used to start the project risk type identification algorithm if there is a risk in the project, and apply the gradient boosting decision tree GBDT algorithm based on the project data to identify the risk category and determine the project risk category; The risk warning management module determines risk response measures based on the risk level and type of the project, matches the project risk database, and pushes them to relevant managers in real time.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the computer program is executed by the processor, the processor is caused to perform the steps of the method according to any one of claims 1 to 5.

Images