Project progress prediction method and system based on deep learning

Abstract

The invention relates to the technical field of deep learning, and discloses a project progress prediction method and system based on deep learning, and the method comprises the steps: obtaining project management data, and converting the project management data into a format of an input project progress prediction model; constructing a project progress prediction model based on deep learning; training a project progress prediction model by using the project management data after format conversion; and predicting a common project by using the trained project progress prediction model to obtain the project progress. Historical project management data, new project management data and real-time project management data are analyzed by using a deep learning technology, key factors influencing project progress are automatically identified, and accurate progress prediction is provided; the accuracy and efficiency of project progress prediction are improved, the method can adapt to constantly changing project environments and conditions, and a more powerful and flexible decision support tool is provided for project management.

Description

Technical Field

[0001] The present invention relates to the field of deep learning technology, and in particular to a project progress prediction method and system based on deep learning. Background Art

[0002] In recent years, artificial intelligence, especially deep learning technology, has made significant progress in many fields. Deep learning technologies, such as neural networks, have been widely used in image recognition, natural language processing and other fields, and have demonstrated excellent complex data processing and analysis capabilities. Applying deep learning technology to project management schedule forecasting is a new field with great potential and is expected to significantly improve the accuracy and efficiency of forecasting.

[0003] Traditional project management schedule forecasting methods, such as rule-based estimation and historical analogy, are effective in some cases, but have limitations when dealing with complex projects. For example, these methods rely too much on personal experience, resulting in subjectivity and inconsistency in forecasts; when dealing with large-scale and multi-dimensional data, traditional methods often lack effective data analysis and processing mechanisms; traditional methods are difficult to adapt to the dynamic changes in project schedules and real-time data updates.

[0004] Existing technologies also have shortcomings in automatically processing and analyzing large amounts of project management data. For example, traditional project management software usually provides static data input and simple algorithms, lacking in-depth analysis of project dynamics and potential risks. These tools fail to fully utilize data analysis technology to provide accurate schedule forecasts. Summary of the invention

[0005] The present invention provides a project progress prediction method and system based on deep learning, which utilizes deep learning technology to analyze historical project management data, new project management data and real-time project management data, automatically identifies key factors affecting project progress, and provides accurate progress prediction; it not only improves the accuracy and efficiency of project progress prediction, but also can adapt to the ever-changing project environment and conditions, providing a more powerful and flexible decision support tool for project management.

[0006] The present invention provides a project progress prediction method based on deep learning, comprising:

[0007] Acquire project management data and transform it into a format that can be input into a project schedule forecasting model;

[0008] Build a project progress prediction model based on deep learning;

[0009] Use the format-converted project management data to train the project progress prediction model;

[0010] The trained project progress prediction model is used to predict ordinary projects to obtain the project progress.

[0011] Preferably, the project management data includes historical project management data, new project management data and real-time project management data; wherein,

[0012] The historical project management data includes detailed progress records and final results of historical projects, budget execution and actual expenditures of historical projects, problems encountered in historical projects and their solutions, resource allocation and utilization of historical projects, and key milestones achieved and timelines of historical projects;

[0013] The new project management data includes the current progress status and phase results of recent projects, the current budget status and expenditure of recent projects, resource allocation and resource consumption data of recent projects, risk assessment and risk management measures of recent projects, and feedback from recent project teams and stakeholders;

[0014] The real-time project management data includes real-time project progress updates and actual progress, real-time data on current budget usage and cost tracking, real-time resource usage and resource availability, real-time project risk monitoring and problem tracking, and real-time team collaboration and communication efficiency data.

[0015] Preferably, converting the format of the project management data into a format for inputting into the project schedule prediction model comprises:

[0016] Take the above data as samples, that is, get m samples, each sample has n project management data, that is, each sample has n features, and the i-th feature of the j-th sample is recorded as

[0017] The feature set of each sample is recorded as in

[0018] Different numerical features are uniformly scaled to a common scale through feature scaling functions;

[0019] The feature scaling function is specifically:

[0020]

[0021] in, represents a normalized data point; Represents a single data point in raw project management data; represents the average value of the above data, that is σ i represents the standard deviation of the above data, that is

[0022] Preferably, the project progress prediction model includes an input layer, a feature extraction layer, a hidden layer and an output layer;

[0023] The input layer receives the historical project management data, new project management data and real-time project management data after format conversion;

[0024] The feature extraction layer automatically identifies key features from historical project management data, new project management data, and real-time project management data using a long short-term memory network;

[0025] The hidden layer obtains the prediction result of the project progress prediction model according to the key features;

[0026] The output layer converts the prediction result of the project progress prediction model into a project progress prediction value.

[0027] Preferably, the long short-term memory network includes a forget gate, an input gate, a unit state update and an output gate;

[0028] The forget gate is used to selectively retain or forget the historical project management data learned from the unit state at the current moment; specifically:

[0029] The output f of the forget gate t Each element of The formula is:

[0030]

[0031] in is a vector, indicating that at the current time t, for the jth sample, the forget gate decides to forget the information for each feature i; σ is the sigmoid function, w f,i is the weight vector of the forget gate for the i-th feature; b f,i is the bias term of the forget gate for the i-th feature, which is used to adjust the activation threshold of the forget gate; Combine the past hidden state with the current input features, is the hidden state at the previous moment, which contains the information learned by the network at the previous moment; The input at the current moment is the historical project management data, new project management data and real-time project management data after format conversion, that is, standardized

[0032] The input gate is used to determine whether new project management data and real-time project management data need to be saved, and to generate unit status updates; specifically, it is expressed by the formula:

[0033]

[0034] Among them, the output of the input gate is a vector representing the i-th feature of the j-th sample at the current time t, which is used to determine the information of new project management data and real-time project management data to be retained at the current time; σ is the sigmoid function, w i,i Represents the weight matrix of dimension 2*n in the input gate, where n is the number of features and the subscript i,i represents the weight of each feature i; Represents the hidden state at the previous moment With the current input splicing; Represents the input at the current moment, which is the historical project management data, new project management data and real-time project management data after format conversion, that is, standardized b i,i Represents the bias term of dimension 1*n in the input gate, which is used to adjust the activation threshold of the forget gate;

[0035] The unit state update includes candidate unit state and updated unit state; wherein the candidate unit state includes new project management data and real-time project management data added to the unit state at the current moment; the updated unit state is the final unit state, including historical information and new information of the candidate unit state; specifically:

[0036] The candidate unit state is represented as:

[0037]

[0038] in, Represents the candidate unit state for the jth sample at the current time t, indicating the new information added to the unit state; Tanh is the hyperbolic tangent activation function, which compresses the value between -1 and 1 and is used to regularize the data; represents a weight matrix of dimension 2*n, where n is the number of features and the subscript c,i represents the weight of each feature i; Represents the hidden state at the previous moment With the current input splicing; Represents the input at the current moment, which is the historical project management data, new project management data and real-time project management data after format conversion, that is, standardized Represents a bias term with dimension 1*n;

[0039] The updated unit state is represented as:

[0040]

[0041] in, Represents the updated unit state for the jth sample at the current time t, which is the information stored in the LSTM network; is the output of the forget gate, indicating that at the current time t, for the jth sample, the forget gate decides how much previous information to forget for each feature i; represents the unit state for j samples at the previous time t-1; is the output of the input gate, representing the i-th feature of the j-th sample at the current time t, which is used to determine the amount of information of the new project management data and the real-time project management data to be retained at the current time; is the candidate cell state at the current moment, representing the new information added to the cell state, which is determined by the current input and the previous hidden state;

[0042] The output gate controls the information flow in the hidden state and determines the final output information transmitted to the current moment; specifically:

[0043] The output of the output gate is expressed as:

[0044]

[0045] in, is the output of the output gate, indicating the amount of information that the output gate decides to retain for feature i at the current time t; is a value between 0 and 1 that determines the slave unit status Passed to the hidden state The amount of information W o,i Represents the weight vector of the output gate for the i-th feature, which determines how the previous hidden state and the current input affect the decision of the output gate; Represents the hidden state of the previous moment, including the information learned so far; is the hidden state at the previous moment and the current input splicing; The input at the current moment is the historical project management data, new project management data and real-time project management data after format conversion, that is, standardized b o,i Represents the bias term of the output gate for feature i, which is used to adjust the activation threshold of the output gate;

[0046] The final output at the current moment is expressed as:

[0047]

[0048] in, is the output at the current time t for feature i, which is composed of the output of the output gate and the unit state at the current time Jointly determined; tanh represents the hyperbolic tangent activation function, which converts the unit state The value of is compressed to between -1 and 1 to regularize the data; Represents the cell state at the current moment, including the accumulated information of the LSTM cell at the current moment.

[0049] Preferably, the hidden layer obtains the prediction result of the project progress prediction model according to the key features, specifically:

[0050] The hidden layer converts the result of the output gate into a predicted value, expressed as:

[0051]

[0052] in, It represents the predicted value of the i-th feature of the j-th sample at the current time t, and the dimension of the predicted value is 1*1; Represents the weight between the predicted value of the i-th feature of the j-th sample at the current time t and the result of the output gate; Represents the output result of the output gate; Represents the bias term of the predicted value of the i-th feature of the j-th sample at the current time t.

[0053] Preferably, the output layer converts the prediction results of the project progress prediction model into a project progress prediction value; specifically including the predicted date of project completion and the progress status of the next stage.

[0054] Preferably, the step of training the project progress prediction model with the format-converted project management data includes:

[0055] Historical project management data, new project management data, and real-time project management data are input, and the weights and biases of the project progress prediction model are optimized using back propagation and gradient descent. The learning rate, batch size, and number of iterations are adjusted to optimize learning efficiency and accuracy.

[0056] Preferably, a project progress prediction system based on deep learning includes:

[0057] A preprocessing module for acquiring project management data and converting it into a format for input into the project schedule forecasting model;

[0058] Model building module, used to build a project progress prediction model based on deep learning;

[0059] Model training module, used to train the project progress prediction model with the project management data after format conversion;

[0060] The progress prediction module is used to use the trained project progress prediction model to predict ordinary projects and obtain the project progress.

[0061] A device comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements a project schedule prediction method based on deep learning when executing the computer program.

[0062] A computer-readable storage medium, the computer-readable storage medium including a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the project progress prediction method based on deep learning.

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] The present invention discloses a project progress prediction method and system based on deep learning, which utilizes deep learning technology to analyze historical project management data, new project management data and real-time project management data, automatically identifies key factors affecting project progress, and provides accurate progress prediction; it not only improves the accuracy and efficiency of project progress prediction, but also can adapt to the ever-changing project environment and conditions, providing a more powerful and flexible decision support tool for project management. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 This is a flowchart of a project progress prediction method based on deep learning provided by an embodiment of the present invention;

[0066] Figure 2 is a management progress prediction flow chart of an elderly education project provided by an embodiment of the present invention;

[0067] Figure 3 is a schematic diagram of a preprocessing process provided by an embodiment of the present invention;

[0068] Figure 4 It is a schematic diagram of the LSTM unit structure provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0069] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0070] In the existing technology, the prediction of project progress usually relies on rule-based estimation or expert experience judgment. These methods often cannot provide high-precision predictions when facing complex and data-intensive projects, and their ability to process large amounts of data is limited. In addition, although existing project management tools provide a certain degree of data processing and analysis functions, they lack deep data mining and learning capabilities and cannot automatically adapt to the dynamic changes of the project, thus limiting the accuracy and practicality of the prediction.

[0071] like Figure 1 As shown, the present application provides a project progress prediction method based on deep learning, including:

[0072] S1: Acquire project management data and convert it into a format that can be input into the project schedule forecasting model;

[0073] S2: Build a project progress prediction model based on deep learning;

[0074] S3: Use the converted project management data to train the project progress prediction model;

[0075] S4: Use the trained project progress prediction model to predict common projects and obtain the project progress.

[0076] Preferably, in step S1, the project management data includes historical project management data, new project management data and real-time project management data; wherein,

[0077] The historical project management data includes detailed progress records and final results of historical projects, budget execution and actual expenditures of historical projects, problems encountered in historical projects and their solutions, resource allocation and utilization of historical projects, and key milestones achieved and timelines of historical projects;

[0078] The new project management data includes the current progress status and phase results of recent projects, the current budget status and expenditure of recent projects, resource allocation and resource consumption data of recent projects, risk assessment and risk management measures of recent projects, and feedback from recent project teams and stakeholders;

[0079] The real-time project management data includes real-time project progress updates and actual progress, real-time data on current budget usage and cost tracking, real-time resource usage and resource availability, real-time project risk monitoring and problem tracking, and real-time team collaboration and communication efficiency data.

[0080] Preferably, converting the format of the project management data into a format for inputting into the project schedule prediction model comprises:

[0081] Take the above data as samples, that is, get m samples, each sample has n project management data, that is, each sample has n features, and the i-th feature of the j-th sample is recorded as

[0082] The feature set of each sample is recorded as in

[0083] Different numerical features are uniformly scaled to a common scale through feature scaling functions;

[0084] The feature scaling function is specifically:

[0085]

[0086] in, represents a normalized data point; Represents a single data point in raw project management data; represents the average value of the above data, that is σ i represents the standard deviation of the above data, that is

[0087] In the above solution, the acquired project management data is converted into a format that can be effectively processed by the deep learning model, including data cleaning, data conversion, data standardization and data segmentation. In the process of data standardization, the feature scaling function is crucial because it helps to achieve comparability between features by scaling different numerical features to a common scale, thereby significantly improving the effect and speed of model training. is converted to a new value centered at 0 with a standard deviation of 1, making the values ​​of different features comparable while preserving the information and structure in the data; after standardization, the data point It reflects the degree of deviation of the original data point from the mean of the entire data set, and the unit of measurement is standard deviation.

[0088] Preferably, in step S2, the project progress prediction model includes an input layer, a feature extraction layer, a hidden layer and an output layer;

[0089] The input layer receives the historical project management data, new project management data and real-time project management data after format conversion;

[0090] The feature extraction layer automatically identifies key features from historical project management data, new project management data, and real-time project management data using a long short-term memory network;

[0091] In the above solution, the feature extraction layer is used to capture long-term dependencies in time series data and automatically identify key features from complex project management data. These features include pattern changes in time series data, efficiency of resource allocation, and key milestones in project progress;

[0092] The hidden layer obtains the prediction result of the project progress prediction model according to the key features;

[0093] The output layer converts the prediction result of the project progress prediction model into a project progress prediction value.

[0094] Preferably, the long short-term memory network includes a forget gate, an input gate, a unit state update and an output gate;

[0095] The forget gate is used to selectively retain or forget the historical project management data learned from the unit state at the current moment; specifically:

[0096] The output f of the forget gate t Each element of The formula is:

[0097]

[0098] in is a vector, indicating that at the current time t, for the jth sample, the forget gate decides to forget the information for each feature i; σ is the sigmoid function, w f,i is the weight vector of the forget gate for the i-th feature; b f,i is the bias term of the forget gate for the i-th feature, which is used to adjust the activation threshold of the forget gate; Combine the past hidden state with the current input features, is the hidden state at the previous moment, which contains the information learned by the network at the previous moment; The input at the current moment is the historical project management data, new project management data and real-time project management data after format conversion, that is, standardized

[0099] In the above scheme, the forget gate indicates that the historical project management data learned before needs to be selectively retained or forgotten from the unit state at the current moment. It may be that the environment of project management may change, and some previous information may be outdated and no longer have the value of predicting future project progress; it may be that some trends or rules in project management have changed recently, and the forget gate can help the model forget those trends that are no longer applicable so as to better adapt to new situations; it may also be that some features have changed a lot in history, but have become relatively stable recently, and the forget gate can help reduce over-reliance on those unstable features. Between 0 and 1, the closer the value is to 1, the more information is retained;

[0100] The input gate is used to determine whether new project management data and real-time project management data need to be saved, and to generate unit status updates; specifically, it is expressed by the formula:

[0101]

[0102] Among them, the output of the input gate is a vector representing the i-th feature of the j-th sample at the current time t, which is used to determine the information of new project management data and real-time project management data to be retained at the current time; σ is the sigmoid function, w i,i Represents the weight matrix of dimension 2*n in the input gate, where n is the number of features and the subscript i,i represents the weight of each feature i; Represents the hidden state at the previous moment With the current input splicing; Represents the input at the current moment, which is the historical project management data, new project management data and real-time project management data after format conversion, that is, standardized b i,i Represents the bias term of dimension 1*n in the input gate, which is used to adjust the activation threshold of the forget gate;

[0103] The unit state update includes candidate unit state and updated unit state; wherein the candidate unit state includes new project management data and real-time project management data added to the unit state at the current moment; the updated unit state is the final unit state, including historical information and new information of the candidate unit state; specifically:

[0104] The candidate unit state is represented as:

[0105]

[0106] in, Represents the candidate unit state for the jth sample at the current time t, indicating the new information added to the unit state; Tanh is the hyperbolic tangent activation function, which compresses the value between -1 and 1 and is used to regularize the data; represents a weight matrix of dimension 2*n, where n is the number of features and the subscript c,i represents the weight of each feature i; Represents the hidden state at the previous moment With the current input splicing; Represents the input at the current moment, which is the historical project management data, new project management data and real-time project management data after format conversion, that is, standardized Represents a bias term with dimension 1*n;

[0107] The updated unit state is represented as:

[0108]

[0109] in, Represents the updated unit state for the jth sample at the current time t, which is the information stored in the LSTM network; is the output of the forget gate, indicating that at the current time t, for the jth sample, the forget gate decides how much previous information to forget for each feature i; represents the unit state for j samples at the previous time t-1; is the output of the input gate, representing the i-th feature of the j-th sample at the current time t, which is used to determine the amount of information of the new project management data and the real-time project management data to be retained at the current time; is the candidate cell state at the current moment, representing the new information added to the cell state, which is determined by the current input and the previous hidden state;

[0110] In the above scheme, the candidate cell state and the update cell state work together to determine how new information is updated to the LSTM cell state. The candidate cell state captures the current progress status, current budget status, resource allocation and other information of the recent project in the new project management data to consider whether to add them to the cell state at the current moment. At the same time, the patterns and trends in the historical project management data are used as the hidden state of the previous moment, so that the model can use past experience to adjust the state at the current moment. This process allows LSTM to maintain long-term dependencies while introducing new important information. By combining the decisions of the forget gate and the input gate to update the cell state, it includes information from the past. Also contains current new information This enables LSTM to effectively maintain information in long sequences and forget irrelevant details when necessary; the cell state can be understood as a unit for storing the current project in project management progress forecasting. The state vector of important information on a specific feature i; it combines historical information with the latest information, which helps the model maintain important key data when predicting project progress, and is passed to the output gate through the hidden state to affect the final prediction output.

[0111] The output gate controls the information flow in the hidden state and determines the final output information transmitted to the current moment; specifically:

[0112] The output of the output gate is expressed as:

[0113]

[0114] in, is the output of the output gate, indicating the amount of information that the output gate decides to retain for feature i at the current time t; is a value between 0 and 1 that determines the slave unit status Passed to the hidden state The amount of information W o,i Represents the weight vector of the output gate for the i-th feature, which determines how the previous hidden state and the current input affect the decision of the output gate; Represents the hidden state of the previous moment, including the information learned so far; is the hidden state at the previous moment and the current input splicing; The input at the current moment is the historical project management data, new project management data and real-time project management data after format conversion, that is, standardized b o,i Represents the bias term of the output gate for feature i, which is used to adjust the activation threshold of the output gate;

[0115] The final output at the current moment is expressed as:

[0116]

[0117] in, is the output at the current time t for feature i, which is composed of the output of the output gate and the unit state at the current time Jointly determined; tanh represents the hyperbolic tangent activation function, which converts the unit state The value of is compressed to between -1 and 1 to regularize the data; Represents the cell state at the current moment, including the accumulated information of the LSTM cell at the current moment.

[0118] In the above scheme, the output gate o t The cell state C is determined by the sigmoid function t How much information needs to be passed to the hidden state h t , and the hidden state h t By adding o t With tanh(C t ) by multiplying the result of . This mechanism allows LSTM to control the flow of information and decide which information is important and should continue to be passed through the network. The output gate determines the importance of feature i in a specific project j at the current time t based on the weights and input information, and how this information is passed to the final output. This enables the model to adaptively and selectively focus on and pass on information related to the project management schedule forecast at the current moment, thereby improving the accuracy of the forecast.

[0119] Preferably, the hidden layer obtains the prediction result of the project progress prediction model according to the key features, specifically:

[0120] The hidden layer converts the result of the output gate into a predicted value, expressed as:

[0121]

[0122] in, It represents the predicted value of the i-th feature of the j-th sample at the current time t, and the dimension of the predicted value is 1*1; Represents the weight between the predicted value of the i-th feature of the j-th sample at the current time t and the result of the output gate; Represents the output result of the output gate; Represents the bias term of the predicted value of the i-th feature of the j-th sample at the current time t.

[0123] Preferably, the output layer converts the prediction results of the project progress prediction model into a project progress prediction value; specifically including the predicted date of project completion and the progress status of the next stage.

[0124] Preferably, in step S3, training the project progress prediction model with the format-converted project management data includes:

[0125] Historical project management data, new project management data, and real-time project management data are input, and the weights and biases of the project progress prediction model are optimized using back propagation and gradient descent. The learning rate, batch size, and number of iterations are adjusted to optimize learning efficiency and accuracy.

[0126] In the above scheme, this application combines deep learning technology and project management data, adopts LSTM deep learning, effectively processes complex time series data, captures long-term dependencies and patterns in the data; can automatically learn from historical and real-time data, and provide data-driven accurate predictions. With the continuous input of new data, the predictive ability of the method is improved, and it can effectively predict the key progress nodes and potential risks of the project, and accurately identify the potential delay patterns of the project. This provides support for project managers to efficiently plan resources and adjust plans in a changing environment.

[0127] Compared with the prior art, this application has significant beneficial effects;

[0128] First, it significantly improves the accuracy of project schedule forecasts by applying deep learning, which enables the model to learn and identify complex patterns that affect project schedules from large amounts of historical and real-time data;

[0129] Secondly, the present invention can effectively process and analyze large amounts of project management data, which is difficult to achieve with traditional methods, especially in complex and data-intensive projects.

[0130] In addition, deep learning models have excellent adaptability and flexibility, and can be continuously optimized over time and with the accumulation of new data, improving the accuracy and applicability of predictions. They can also improve project management efficiency and help project managers plan resources and adjust plans more effectively through accurate predictions.

[0131] Finally, by identifying potential project delays and risks in advance, the additional costs and resource waste caused by project delays are reduced, and the risk of project failure is reduced. These advantages make the present invention have important practical application value in projects of various sizes and types, especially complex and data-intensive projects.

[0132] Preferably, a project progress prediction system based on deep learning includes:

[0133] A preprocessing module for acquiring project management data and converting it into a format for input into the project schedule forecasting model;

[0134] Model building module, used to build a project progress prediction model based on deep learning;

[0135] The model training module is used to train the project progress prediction model with the format-converted project management data, tune the loss function and optimizer, and update the model parameters through the back-propagation algorithm;

[0136] The progress prediction module is used to use the trained project progress prediction model to predict ordinary projects and obtain the project progress.

[0137] In one embodiment provided in the present application, a method for predicting the progress of elderly education project management based on deep learning is provided. In order to better illustrate the effectiveness and practicality of the present invention, we will describe a specific application case: progress management of elderly education projects.

[0138] like Figure 2 As shown, it is a flow chart of management progress prediction of elderly education project provided by an embodiment of the present invention, and the method mainly includes the following steps:

[0139] Step 1: In this embodiment of the present invention, data source: the data set is collected from similar elderly education projects in the past five years, including the following types of project data:

[0140] Historical project management data: detailed progress records and final results of completed projects, budget execution and actual expenditures of past projects, problems encountered in historical projects and their solutions, resource allocation and utilization of past projects, key milestones achieved and timelines of previous projects;

[0141] New project management data: current progress status and phase results of recent projects, current budget status and expenditure of recent projects, latest resource allocation and resource consumption data, risk assessment and risk management measures of current projects, latest feedback from project teams and stakeholders;

[0142] Real-time project management data: real-time project progress updates and actual progress, real-time data on current budget usage and cost tracking, real-time resource usage and resource availability, real-time project risk monitoring and issue tracking, real-time team collaboration and communication efficiency data. This data can be obtained from project management software and databases, meeting minutes and documents, financial reports, real-time project tracking tools, risk assessment software, feedback systems, etc.

[0143] In this embodiment of the present invention, a total of 5 samples are obtained, i.e., m=5, which are similar elderly education projects in the past five years. Each sample has 8 characteristics, i.e., n=8, which are: course design days, teaching material preparation cost, teacher training feedback score, student satisfaction score, technical support time, marketing promotion expenses, project management team satisfaction, and online platform availability score.

[0144] like Figure 3As shown, the obtained project management data needs to be preprocessed. First, data cleaning is performed: remove or correct erroneous records in historical and new project data. In the embodiment of the present invention, there are no erroneous records; handle missing values: for missing budget data, use the average value of similar projects to fill. In the embodiment of the present invention, there are no missing values; remove duplicate records to ensure the uniqueness of the data. Next is data conversion: text data conversion: convert problem descriptions and feedback into quantifiable values; time format unification: convert all date and time data into a unified date format (for example, YYYY-MM-DD); encode categorical data: for example, convert project stages (such as "planning", "execution", "end") into digital codes (0, 1, 2). The data in the embodiment of the present invention are all continuous or ordered categorical features and can be used directly. Then comes data standardization: use standard deviation to standardize time and cost data. For each feature, calculate its average value (μ i ) and standard deviation (σ i ), and then apply the formula Obtain standardized features Finally, data segmentation is performed: the data is divided into a training set and a test set, for example, 80% of the data is used as the training set and the remaining 20% ​​is used as the test set. The following is a data set of an embodiment of the present invention, and has been standardized.

[0145]

[0146]

[0147] Step 2: Based on the preprocessed data, a deep learning model for project management progress prediction is constructed, including designing a deep learning model suitable for project management progress prediction, a model architecture, and initialization parameters. In the embodiment of the present invention, the Python programming language and the TensorFlow framework are used to construct the model. The TensorFlow library, the Sequential model class in the Keras sub-library of TensorFlow, and the LSTM and Dense in the Keras sub-library of TensorFlow are imported.

[0148] like Figure 4 As shown, the model architecture of the elderly education project is defined:

[0149] Input layer: Receives preprocessed project management data features, including course design days, teaching material preparation costs, teacher training feedback scores, student satisfaction scores, technical support time, marketing costs, project management team satisfaction, and online platform usability scores, a total of 8 data features.

[0150] Feature extraction layer: Enhance the expressiveness and learning ability of the model through multi-layer stacking; enable the model to process and learn more complex relationships and patterns in the data, and improve the accuracy and generalization of predictions. In an embodiment of the present invention, two feature extraction layers are initially added to process time series data, and each layer has 16 units to process time series data. After setting the initial number of layers and units, continuous experiments are required to adjust the number of layers and units to determine the best model. The parameters involved in the entire deep learning model require continuous experiments and hyperparameter tuning to obtain the best performing model.

[0151] Hidden layer: Three fully connected layers (i.e., three hidden layers, and these hidden layers are all fully connected) are added after the LSTM layer. Each fully connected layer contains 128 neurons, and ReLU is used as the activation function. These layers can further process and integrate the features extracted by the LSTM layer. The number of hidden layers and the number of neurons are hyperparameters that need to be adjusted. In an embodiment of the present invention, three fully connected layers and 128 neurons in each layer are selected considering the common and effective network architecture. Such a setting can not only capture complex patterns in the data, but also avoid overfitting problems caused by overly complex models. In addition, ReLU is used as an activation function because it is simple to calculate and can alleviate the problem of gradient disappearance, which helps the training of the network. These layers can further process and integrate the features extracted by the LSTM layer to achieve more accurate project management progress prediction.

[0152] Output layer: The output layer is a single neuron, which predicts the total number of days required to complete the project. Then instantiate the model and define the architecture, define the loss function and optimizer. The loss function is used to quantify the difference between the predicted results of the model and the actual results. In the embodiment of the present invention, since the project management progress forecast is a prediction of continuous values, MSE is a commonly used regression loss function, which is very suitable for the prediction of continuous values; MSE can quantify the difference between the predicted results of the model and the actual results, reflecting the prediction accuracy of the model; it can be minimized by optimization algorithms such as gradient descent, thereby improving the performance of the model. Therefore, in the embodiment of the present invention, the mean square error MSE is specified as the loss function. in is the true value, is the predicted value, n is the number of samples, and the smaller the loss value is, the closer the model's prediction is to the true value. In an embodiment of the present invention, project management progress forecasting usually involves processing time series data, and the characteristic of this type of data is that there is a temporal dependency between observation points. The Adam optimizer can effectively handle this dependency because it can adaptively adjust the learning rate so that the model can better capture the dynamic changes in the data, and project management data usually contains multiple related features, such as the scale and complexity of the project, the skills of team members, etc. The Adam optimizer can assign a different learning rate to each parameter, so that the model can effectively handle high-dimensional features. Therefore, the Adam optimizer is specified in an embodiment of the present invention.

[0153] Step 3: Use the preprocessed data to train the deep learning model for project management progress prediction, tune the loss function and optimizer, and update the model parameters through the back propagation algorithm so that the model can accurately reflect the project management progress. During the training process, ensure that the input data The format and dimensions of the and labels match the model input and output. The data is first forward propagated through the layers of the model. At each LSTM and fully connected layer, the model uses the current weights and biases to calculate the output The output of the model is compared with the true value, and the loss value is calculated, which is a quantitative representation of the prediction accuracy of the model. According to the loss value, the gradient of the weights and biases of each layer is calculated by the back propagation algorithm. The gradient represents the derivative of the loss function with respect to the parameter, indicating the direction in which the loss function increases or decreases. Use the Adam optimizer to update the weights and biases of the model. The performance of the model is evaluated using the test set previously segmented. Adjust the hyperparameters according to the performance of the model on the test data, such as increasing or decreasing the number of neurons, changing the learning rate, or adjusting the number of layers. The training process of the deep learning model is iterative, and the architecture and hyperparameters of the model need to be adjusted multiple times. In an embodiment of the present invention, the data set is divided into 80% training set and 20% test set. Set the batch size to 5 and the number of iterations (Epochs) to 500.

[0154] Step 4: Use the constructed deep learning model for project management progress prediction to make predictions. Load the trained and saved model, and use the same preprocessing method as the training data to process the new project management data. Use the model to make predictions on the new project management data. This can be achieved by calling the predict method of the model. Convert the prediction results of the model into actual project progress information. In this embodiment of the present invention, the total number of days required to complete the project is calculated using the constructed deep model for project management progress prediction. The prediction result can be directly used as an estimate, expressed as

[0155] In the implementation of this application, the performance of the model on the test set is as follows: Loss value (MSE): The MSE on the test set is 0.025, showing that the model has a high accuracy in predicting indicators related to project management. Accuracy: In certain specific prediction tasks, such as predicting project completion time, the model achieved an accuracy of about 90%. This model training process shows that it is feasible to use deep learning methods, especially LSTM, to predict different aspects of elderly education project management. The model performs well in processing complex time series data and can provide valuable insights and predictions, thereby helping project management teams improve the quality of decision-making.

[0156] The deep learning-based intelligent prediction method for the progress of elderly education project management provided in the embodiment of the present application realizes accurate prediction of the progress of elderly education project management through a series of steps, proving that the method can provide accurate progress prediction, thereby improving the accuracy and efficiency of project progress prediction, and providing strong support for project success. This application is not only applicable to project management in the field of elderly education, but also particularly applicable to those complex and data-intensive project environments, marking an important progress in the intelligent and precise aspects of project management methods.

[0157] A device comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements a project schedule prediction method based on deep learning when executing the computer program.

[0158] A computer-readable storage medium, the computer-readable storage medium including a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the project progress prediction method based on deep learning.

[0159] Preferably, the computer program can be divided into one or more modules / units (such as computer program, computer program), and the one or more modules / units are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units can be a series of computer program instruction segments that can complete specific functions, and the instruction segments are used to describe the execution process of the computer program in the terminal device.

[0160] The processor may be a central processing unit (CPU), or other general-purpose processors, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor, or the processor may be any conventional processor. The processor is the control center of the terminal device, and various parts of the terminal device are connected using various interfaces and lines.

[0161] The memory mainly includes a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function, etc., and the data storage area can store related data, etc. In addition, the memory can be a high-speed random access memory, or a non-volatile memory, such as a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, and a flash card (Flash Card), etc., or the memory can also be other volatile solid-state storage devices.

[0162] It should be noted that the above-mentioned terminal device may include, but is not limited to, a processor and a memory. Those skilled in the art will understand that the above-mentioned terminal device is merely an example and does not constitute a limitation on the terminal device. It may include more or fewer components, or a combination of certain components, or different components.

[0163] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A project progress prediction method based on deep learning, characterized in that: include: Acquire project management data and transform it into a format that can be input into a project schedule forecasting model; Build a project progress prediction model based on deep learning; Use the format-converted project management data to train the project progress prediction model; The trained project progress prediction model is used to predict ordinary projects to obtain the project progress.

2. A project progress prediction method based on deep learning according to claim 1, characterized in that: The project management data includes historical project management data, new project management data and real-time project management data; wherein, The historical project management data includes detailed progress records and final results of historical projects, budget execution and actual expenditures of historical projects, problems encountered in historical projects and their solutions, resource allocation and utilization of historical projects, and key milestones achieved and timelines of historical projects; The new project management data includes the current progress status and phase results of recent projects, the current budget status and expenditure of recent projects, resource allocation and resource consumption data of recent projects, risk assessment and risk management measures of recent projects, and feedback from recent project teams and stakeholders; The real-time project management data includes real-time project progress updates and actual progress, real-time data on current budget usage and cost tracking, real-time resource usage and resource availability, real-time project risk monitoring and problem tracking, and real-time team collaboration and communication efficiency data.

3. A project progress prediction method based on deep learning according to claim 2, characterized in that: The format of project management data is converted into the format for input into the project schedule forecasting model including: Take the above data as samples, that is, get m samples, each sample has n project management data, that is, each sample has n features, and the i-th feature of the j-th sample is recorded as 1≤i≤n; 1≤j≤m; The feature set of each sample is recorded as in 1≤j≤m; Different numerical features are uniformly scaled to a common scale through feature scaling functions; The feature scaling function is specifically: in, represents a normalized data point; Represents a single data point in raw project management data; represents the average value of the above data, that is σ i represents the standard deviation of the above data, that is 4. A project progress prediction method based on deep learning according to claim 3, characterized in that: The project progress prediction model includes an input layer, a feature extraction layer, a hidden layer and an output layer; The input layer receives the historical project management data, new project management data and real-time project management data after format conversion; The feature extraction layer automatically identifies key features from historical project management data, new project management data, and real-time project management data using a long short-term memory network; The hidden layer obtains the prediction result of the project progress prediction model according to the key features; The output layer converts the prediction result of the project progress prediction model into a project progress prediction value.

5. A project progress prediction method based on deep learning according to claim 4, characterized in that: The long short-term memory network includes a forget gate, an input gate, a unit state update and an output gate; The forget gate is used to selectively retain or forget the historical project management data learned from the unit state at the current moment; specifically: The output f of the forget gate t Each element of The formula is: in is a vector, indicating that at the current time t, for the jth sample, the forget gate decides to forget the information for each feature i; σ is the sigmoid function, w f,i is the weight vector of the forget gate for the i-th feature; b f,i is the bias term of the forget gate for the i-th feature, which is used to adjust the activation threshold of the forget gate; Combine the past hidden state with the current input features, is the hidden state at the previous moment, which contains the information learned by the network at the previous moment; The input at the current moment is the historical project management data, new project management data and real-time project management data after format conversion, that is, standardized 1≤i≤n; 1≤j≤m; The input gate is used to determine whether new project management data and real-time project management data need to be saved, and to generate unit status updates; specifically, it is expressed by the formula: Among them, the output of the input gate is a vector representing the i-th feature of the j-th sample at the current time t, which is used to determine the information of new project management data and real-time project management data to be retained at the current time; σ is the sigmoid function, w i,i represents the weight matrix of dimension 2*n in the input gate, where n is the number of features and the subscript i,i represents the weight of each feature i; Represents the hidden state at the previous moment With the current input splicing; Represents the input at the current moment, which is the historical project management data, new project management data and real-time project management data after format conversion, that is, standardized 1≤i≤n; 1≤j≤m; b i,i Represents the bias term of dimension 1*n in the input gate, which is used to adjust the activation threshold of the forget gate; The unit state update includes candidate unit state and updated unit state; wherein the candidate unit state includes new project management data and real-time project management data added to the unit state at the current moment; the updated unit state is the final unit state, including historical information and new information of the candidate unit state; specifically: The candidate unit state is represented as: in, Represents the candidate unit state for the jth sample at the current time t, indicating the new information added to the unit state; Tanh is the hyperbolic tangent activation function, which compresses the value between -1 and 1 and is used to regularize the data; represents a weight matrix of dimension 2*n, where n is the number of features and the subscript c,i represents the weight of each feature i; Represents the hidden state at the previous moment With the current input splicing; Represents the input at the current moment, which is the historical project management data, new project management data and real-time project management data after format conversion, that is, standardized 1≤i≤n; 1≤j≤m; Represents a bias term with dimension 1*n; The updated unit state is represented as: in, Represents the updated unit state for the jth sample at the current time t, which is the information stored in the LSTM network; is the output of the forget gate, indicating that at the current time t, for the jth sample, the forget gate decides how much previous information to forget for each feature i; represents the unit state for j samples at the previous time t-1; is the output of the input gate, representing the i-th feature of the j-th sample at the current time t, which is used to determine the amount of information of the new project management data and the real-time project management data to be retained at the current time; is the candidate cell state at the current moment, representing the new information added to the cell state, which is determined by the current input and the previous hidden state; The output gate controls the information flow in the hidden state and determines the final output information transmitted to the current moment; specifically: The output of the output gate is expressed as: in, is the output of the output gate, indicating the amount of information that the output gate decides to retain for feature i at the current time t; is a value between 0 and 1 that determines the slave unit status Passed to the hidden state The amount of information W o,i Represents the weight vector of the output gate for the i-th feature, which determines how the previous hidden state and the current input affect the decision of the output gate; Represents the hidden state of the previous moment, including the information learned so far; is the hidden state at the previous moment and the current input splicing; The input at the current moment is the historical project management data, new project management data and real-time project management data after format conversion, that is, standardized 1≤i≤n; 1≤j≤m; b o,i Represents the bias term of the output gate for feature i, which is used to adjust the activation threshold of the output gate; The final output at the current moment is expressed as: in, is the output at the current time t for feature i, which is composed of the output of the output gate and the unit state at the current time Jointly determined; tanh represents the hyperbolic tangent activation function, which converts the unit state The value of is compressed to between -1 and 1 to regularize the data; Represents the cell state at the current moment, including the accumulated information of the LSTM cell at the current moment.

6. A project progress prediction method based on deep learning according to claim 5, characterized in that: The hidden layer obtains the prediction results of the project progress prediction model according to the key features, specifically: The hidden layer converts the result of the output gate into a predicted value, expressed as: in, It represents the predicted value of the i-th feature of the j-th sample at the current time t, and the dimension of the predicted value is 1*1; Represents the weight between the predicted value of the i-th feature of the j-th sample at the current time t and the result of the output gate; Represents the output result of the output gate; Represents the bias term of the predicted value of the i-th feature of the j-th sample at the current time t.

7. A project progress prediction method based on deep learning according to claim 6, characterized in that: The output layer converts the prediction results of the project progress prediction model into project progress prediction values, including the predicted date of project completion and the progress status of the next stage.

8. A project progress prediction method based on deep learning according to claim 7, characterized in that: The training of the project progress prediction model using the format-converted project management data includes: Historical project management data, new project management data, and real-time project management data are input, and the weights and biases of the project progress prediction model are optimized using back propagation and gradient descent. The learning rate, batch size, and number of iterations are adjusted to optimize learning efficiency and accuracy.

9. A project progress prediction system based on deep learning, characterized in that: include: A preprocessing module for acquiring project management data and converting it into a format for input into the project schedule forecasting model; Model building module, used to build a project progress prediction model based on deep learning; Model training module, used to train the project progress prediction model with the project management data after format conversion; The progress prediction module is used to use the trained project progress prediction model to predict ordinary projects and obtain the project progress.

Images