Dynamic event association degree quantitative modeling method and system

Abstract

The invention belongs to the technical field of project management, and relates to a dynamic event association degree quantitative modeling method and system. The method comprises the following steps: carrying out structured processing on a recorded text, and carrying out vectorization expression; identifying the event and extracting time information of the event; according to the semantic similarity, the time closeness degree and the event dependency degree between the events, the event association degree is determined; adjusting the priority and reminding time of the event by using a dynamic programming algorithm; calculating the completion probability of each node of the event, and predicting the completion probability of the event; and adjusting an event reminding strategy according to feedback data of the user by adopting a self-adaptive learning algorithm. According to the method, event information can be automatically extracted from daily flow records, quantitative modeling is carried out based on dynamic event association degree, the priority of events is adjusted by using a dynamic planning algorithm, an event management strategy and reminding time can be intelligently adjusted along with the progress of the events, and intelligent association, progress tracking and reminding optimization of the events are realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of project management, and more specifically, relates to a method and system for quantifying and modeling the correlation degree of dynamic events. Background Art

[0002] Current project management tools usually rely on users to manually input events, time nodes, and correlation relationships, lacking intelligent data extraction and automatic management capabilities. Although some existing systems use natural language processing technology for text analysis, there are still deficiencies in accurately extracting events, times, and related events, mainly manifested in:

[0003] It is unable to accurately distinguish between background information and actual event information in daily records.

[0004] The correlation degree between events is difficult to measure, resulting in insufficient accuracy of event automatic management.

[0005] Existing systems lack intelligent reminder and dynamic update capabilities and cannot adapt to changes in projects. Summary of the Invention

[0006] To solve the above technical problems, the present invention provides a method and system for quantifying and modeling the correlation degree of dynamic events.

[0007] In a first aspect, the present invention provides a method for quantifying and modeling the correlation degree of dynamic events, including:

[0008] Structurally process the recorded text and represent it in vector form;

[0009] Identify events and extract the time information of the events;

[0010] Calculate the semantic similarity between events, and at the same time calculate the time proximity between events according to the time information of the events;

[0011] Assign weights to events according to the dependency relationship between events to obtain the event dependency degree between events;

[0012] Determine the event correlation degree according to the semantic similarity, time proximity, and event dependency degree between events;

[0013] Adjust the priority and reminder time of events using the dynamic programming algorithm according to the event correlation degree;

[0014] Construct a Bayesian network based on the event stream, calculate the probability of completion of each node of the event, and predict the probability of event completion;

[0015] Adopt an adaptive learning algorithm to adjust the event reminder strategy according to the feedback data of users.

[0016] In a second aspect, the present invention provides a dynamic event correlation degree quantification and modeling system, including a text processing unit, an identification and extraction unit, a first analysis and processing unit, a second analysis and processing unit, a data processing unit, a first adjustment unit, a prediction unit, and a second adjustment unit;

[0017] The text processing unit is used for structurally processing the recorded text and performing vectorization representation;

[0018] The identification and extraction unit is used for identifying events and extracting the time information of the events;

[0019] The first analysis and processing unit is used for calculating the semantic similarity between events and calculating the time proximity between events according to the time information of the events;

[0020] The second analysis and processing unit is used for assigning weights to events according to the dependency relationship between events to obtain the event dependency between events;

[0021] The data processing unit is used for determining the event correlation degree according to the semantic similarity, time proximity, and event dependency between events;

[0022] The first adjustment unit is used for adjusting the priority and reminder time of events by using a dynamic programming algorithm according to the event correlation degree;

[0023] The prediction unit is used for constructing a Bayesian network based on the event stream, calculating the probabilities of completion of each node of the event, and predicting the probability of event completion;

[0024] The second adjustment unit is used for adopting an adaptive learning algorithm to adjust the event reminder strategy according to the user's feedback data.

[0025] Based on the above technical solutions, the present invention can also be improved as follows.

[0026] Further, a natural language processing algorithm is adopted to structurally process the recorded text, including:

[0027] Performing word segmentation on the recorded text, annotating according to the part of speech, identifying entities in the recorded text, extracting key information from the recorded text, and organizing it according to a preset structural format;

[0028] Extracting the mapping relationship between the text features and structural information of the key information, constructing a machine learning model and performing model training, and using the machine learning model to structurally process the key information of the recorded text.

[0029] Further, a pre-trained language model is used to perform vectorization representation on the recorded text.

[0030] Further, the identification of events includes:

[0031] Describe events using a triple structure; the triple structure includes time information, action information, and objects;

[0032] Build a deep learning model, including a BERT model, a bidirectional long short-term memory network, and a conditional random field layer;

[0033] Collect text data and perform manual annotation on the text data related to events;

[0034] Divide the annotated text data into a training set, a validation set, and a test set;

[0035] Load the pre-trained BERT model and convert the text data into a sequence of word vectors;

[0036] The BERT model performs bidirectional context feature extraction on the input text to capture semantic information in the text;

[0037] Input the feature vectors output by the BERT model into the bidirectional long short-term memory network. The bidirectional long short-term memory network extracts the semantic features of the text and outputs the hidden layer state;

[0038] Use the output of the bidirectional long short-term memory network as the input to the conditional random field layer. The conditional random field layer constrains and optimizes the prediction results of the bidirectional long short-term memory network to obtain the final event extraction result;

[0039] Use the training set to train the deep learning model and adjust the parameters of the model by minimizing the loss function;

[0040] Use the trained deep learning model to obtain the event recognition result.

[0041] Furthermore, calculate the semantic similarity between events, including: calculating the cosine similarity of the vectors corresponding to the events to obtain the semantic similarity between events.

[0042] Furthermore, use an exponential decay function to calculate the temporal proximity between events.

[0043] Furthermore, determine the event correlation degree based on the semantic similarity, temporal proximity, and event dependency between events, including:

[0044] Let the event correlation degree be S event , the semantic similarity between events be S semantic , the temporal proximity between events be S time , the event dependency between events be S dependency , the weight of the semantic similarity be α, the weight of the temporal proximity be β, and the weight of the event dependency be γ, then:

[0045] S event= αS semantic + βS time + γS dependency 。

[0046] Furthermore, according to the event correlation degree, use the dynamic programming algorithm to adjust the priority and reminder time of the event, including:

[0047] Take the priority of the event as the state variable, set the completion time of the event and the resource usage as the benefit parameters, construct the state transition equation, and calculate the state variable corresponding to the maximum benefit; set the time at a set duration from the deadline as the reminder time.

[0048] Furthermore, adopt the adaptive learning algorithm to adjust the event reminder strategy according to the user's feedback data, including:

[0049] Collect the event reminder strategies and the user's feedback data in the historical events; the user's feedback data includes the event creation time, the event completion time, and the event priority; the event reminder strategy includes the reminder time and the reminder method;

[0050] Build an event reminder model based on the machine learning algorithm. The event reminder model extracts feature data from the user's feedback data, takes the reminder time and the reminder method as the output of the event reminder model, trains the event reminder model, updates the parameters of the event reminder model, and obtains the target event reminder model;

[0051] Input the user's feedback data corresponding to the target event into the target event reminder model to obtain the event reminder strategy of the target event.

[0052] The beneficial effects of the present invention are as follows: The present invention determines the event correlation degree according to the semantic similarity, time proximity, and event dependence between events, can automatically extract event information from the daily running records, and based on the dynamic event correlation degree for quantitative modeling, uses the dynamic programming algorithm to adjust the priority of the event, can intelligently adjust the event management strategy and reminder time according to the event progress, and realizes the intelligent association, progress tracking, and reminder optimization of the event. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is the schematic diagram of the method for quantitative modeling of dynamic event correlation degree provided by Embodiment 1 of the present invention;

[0054] Figure 2 It is the system block diagram of the system for quantitative modeling of dynamic event correlation degree provided by Embodiment 2 of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0055] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. The components of the embodiments of the present invention usually described and illustrated in the drawings here can be arranged and designed in various different configurations.

[0056] Embodiment 1

[0057] As an embodiment, as shown in the appended Figure 1 figures, to solve the above technical problems, this embodiment provides a method for quantifying and modeling dynamic event correlation, including:

[0058] Performing structured processing on the recorded text and performing vectorization representation;

[0059] Identifying events and extracting the time information of the events;

[0060] Calculating the semantic similarity between events, and at the same time calculating the time proximity between events according to the time information of the events;

[0061] Assigning weights to events according to the dependency relationship between events to obtain the event dependency between events;

[0062] Determining the event correlation according to the semantic similarity, time proximity, and event dependency between events;

[0063] Adjusting the priorities and reminder times of events using the dynamic programming algorithm according to the event correlation;

[0064] Constructing a Bayesian network based on the event stream, calculating the probabilities of completion of each node of the event, and predicting the probability of event completion;

[0065] Adopting an adaptive learning algorithm to adjust the event reminder strategy according to the feedback data of the user.

[0066] The present invention determines the event correlation according to the semantic similarity, time proximity, and event dependency between events, can automatically extract event information from daily flow records, and based on the quantification and modeling of dynamic event correlation, uses the dynamic programming algorithm to adjust the priorities of events, and can intelligently adjust the event management strategy and reminder time according to the event progress, realizing intelligent association, progress tracking, and reminder optimization of events.

[0067] Optionally, a natural language processing algorithm is used to perform structured processing on the recorded text, including:

[0068] Performing word segmentation on the recorded text, annotating according to the part of speech, identifying entities in the recorded text, extracting key information from the recorded text, and organizing it according to a preset structured format;

[0069] Extract the mapping relationship between the text features of key information and the structured information, construct a machine learning model and perform model training, and use the machine learning model to perform structured processing on the key information of the recorded text.

[0070] Optionally, use a pre-trained language model to perform vector representation on the recorded text.

[0071] Use a pre-trained language model to perform vector representation on the text to improve the accuracy of event recognition. Examples of pre-trained language models include the BERT (Bidirectional Encoder Representations from Transformers) model and the GPT (Generative Pretrained Transformer) model.

[0072] Optionally, perform event recognition, including:

[0073] Describe the event using a triple structure; the triple structure includes time information, action information, and object; let the time information be T, the action information be A, and the object be O, then the triple is represented as E=(T, A, O);

[0074] Construct a deep learning model, including the BERT model, bidirectional long short-term memory network, and conditional random field layer;

[0075] Collect text data and perform manual annotation on the text data related to the event;

[0076] Divide the annotated text data into a training set, a validation set, and a test set;

[0077] Load the pre-trained BERT model and convert the text data into a sequence of word vectors;

[0078] The BERT model performs bidirectional context feature extraction on the input text to capture the semantic information in the text;

[0079] Input the feature vectors output by the BERT model into the bidirectional long short-term memory network. The bidirectional long short-term memory network extracts the semantic features of the text and outputs the hidden layer state;

[0080] Use the output of the bidirectional long short-term memory network as the input of the conditional random field layer. The conditional random field layer constrains and optimizes the prediction results of the bidirectional long short-term memory network to obtain the final event extraction result;

[0081] Use the training set to train the deep learning model and adjust the parameters of the model by minimizing the loss function;

[0082] Use the trained deep learning model to obtain the event recognition result.

[0083] Action information such as submitting reports, holding meetings, and module development, and objects such as requirement documents and design files.

[0084] Manually annotate to add correct labels to the text data, obtain the corresponding relationship between the text data and the events, thereby improving the performance and accuracy of the model. The training set is used for training the deep learning model to learn the features and patterns of the data; the validation set is used to evaluate the performance of the deep learning model during training, adjust the hyperparameters of the model, and prevent overfitting; the test set is used to evaluate the final performance of the deep learning model after the training of the deep learning model is completed.

[0085] The pre-trained BERT model can capture the bidirectional context information of the text and enhance the accuracy of semantic understanding. The bidirectional long short-term memory network can process sequence data, learn the features of the sequence from the forward and backward directions, and further extract the semantic features of the text. The conditional random field layer then constrains and optimizes the prediction results of the bidirectional long short-term memory network, takes into account the dependencies between the labels, and improves the accuracy of the prediction. By constructing a deep learning model, data can be effectively processed, the semantic information and sequence features of the text can be captured, and the accuracy of event recognition.

[0086] During the training process, input the training set into the deep learning model constructed by the BERT model, the bidirectional long short-term memory network, and the conditional random field layer, calculate the difference between the prediction result of the deep learning model and the true label, and adjust the parameters of the deep learning model by minimizing the loss function (such as the cross-entropy loss function) to ensure the accuracy of the prediction result of the deep learning model.

[0087] Optionally, calculate the semantic similarity between events, including: calculating the cosine similarity of the vectors corresponding to the events to obtain the semantic similarity between the events.

[0088] Optionally, use the exponential decay function to calculate the temporal proximity between events.

[0089] Let t i be the time point of the i-th event, t j be the time point of the j-th event, and the temporal proximity between the events be S time , and λ be the time decay coefficient, then:

[0090] Optionally, determine the event correlation degree according to the semantic similarity, temporal proximity, and event dependency between events, including:

[0091] Let the event correlation degree be S event, the semantic similarity between events is S semantic , the temporal proximity between events is S time , the event dependency between events is S dependency , the weight of semantic similarity is α, the weight of temporal proximity is β, and the weight of event dependency is γ, then:

[0092] S event = αS semantic + βS time + γS dependency .

[0093] Optionally, according to the event correlation degree, use the dynamic programming algorithm to adjust the priority and reminder time of events, including:

[0094] Take the priority of the event as the state variable, set the completion time of the event and the resource usage as the benefit parameters, construct the state transition equation, and calculate the state variable corresponding to the maximum benefit; set the time at a set duration from the deadline as the reminder time.

[0095] Optionally, adopt an adaptive learning algorithm to adjust the event reminder strategy according to the user's feedback data, including:

[0096] Collect the event reminder strategies and the user's feedback data in historical events; the user's feedback data includes the event creation time, event completion time, and event priority; the event reminder strategies include the reminder time and reminder method; the reminder time such as the time of creating a meeting event, the actual completion time of the meeting event, and the priority set for the meeting event. The reminder methods such as SMS reminder and application notification;

[0097] Build an event reminder model based on the machine learning algorithm. The event reminder model extracts feature data from the user's feedback data, takes the reminder time and reminder method as the output of the event reminder model, trains the event reminder model, updates the parameters of the event reminder model, and obtains the target event reminder model;

[0098] Input the user's feedback data corresponding to the target event into the target event reminder model to obtain the event reminder strategy of the target event.

[0099] The present invention can be widely applied to fields such as enterprise project management, personal affairs management, engineering construction management, etc., is applicable to scenarios that require automatic recording, analysis, and management of event progress, can automatically extract event information from daily transaction records, and based on the dynamic event correlation degree, perform quantitative modeling to achieve intelligent association, progress tracking, and reminder optimization of events, and provide an intelligent solution for project management.

[0100] Embodiment 2

[0101] Based on the same principle as the method shown in Embodiment 1 of the present invention, as shown in the appendix Figure 2 There is also provided a dynamic event correlation degree quantification modeling system, including a text processing unit, an identification and extraction unit, a first analysis and processing unit, a second analysis and processing unit, a data processing unit, a first adjustment unit, a prediction unit, and a second adjustment unit;

[0102] The text processing unit is used to perform structured processing on the recorded text and perform vectorization representation;

[0103] The identification and extraction unit is used to identify events and extract the time information of the events;

[0104] The first analysis and processing unit is used to calculate the semantic similarity between events and calculate the time proximity between events according to the time information of the events;

[0105] The second analysis and processing unit is used to assign weights to events according to the dependency relationship between events to obtain the event dependency degree between events;

[0106] The data processing unit is used to determine the event correlation degree according to the semantic similarity, time proximity, and event dependency degree between events;

[0107] The first adjustment unit is used to adjust the priority and reminder time of events by using the dynamic programming algorithm according to the event correlation degree;

[0108] The prediction unit is used to construct a Bayesian network based on the event stream, calculate the probability of completion of each node of the event, and predict the probability of event completion;

[0109] The second adjustment unit is used to adopt an adaptive learning algorithm to adjust the event reminder strategy according to the user's feedback data.

[0110] Optionally, a natural language processing algorithm is used to perform structured processing on the recorded text, including:

[0111] Segment the recorded text, annotate according to the part of speech, identify the entities in the recorded text, extract the key information from the recorded text, and organize it according to the preset structured format;

[0112] Extract the mapping relationship between the text features and structured information of the key information, construct a machine learning model and perform model training, and use the machine learning model to perform structured processing on the key information of the recorded text.

[0113] Optionally, a pre-trained language model is used to perform vectorization representation on the recorded text.

[0114] Optionally, the identification of events includes:

[0115] Describe events using a triple structure; the triple structure includes time information, action information, and objects;

[0116] Construct a deep learning model, including a BERT model, a bidirectional long short-term memory network, and a conditional random field layer;

[0117] Collect text data and perform manual annotation on the text data related to events;

[0118] Divide the annotated text data into a training set, a validation set, and a test set;

[0119] Load the pre-trained BERT model and convert the text data into a sequence of word vectors;

[0120] The BERT model performs bidirectional context feature extraction on the input text to capture semantic information in the text;

[0121] Input the feature vectors output by the BERT model into the bidirectional long short-term memory network. The bidirectional long short-term memory network extracts the semantic features of the text and outputs the hidden layer state;

[0122] Use the output of the bidirectional long short-term memory network as the input to the conditional random field layer. The conditional random field layer constrains and optimizes the prediction results of the bidirectional long short-term memory network to obtain the final event extraction result;

[0123] Use the training set to train the deep learning model and adjust the parameters of the model by minimizing the loss function;

[0124] Obtain the event recognition result using the trained deep learning model.

[0125] Optionally, calculate the semantic similarity between events, including: calculating the cosine similarity of the vectors corresponding to the events to obtain the semantic similarity between events.

[0126] Optionally, use an exponential decay function to calculate the temporal proximity between events.

[0127] Optionally, determine the event correlation based on the semantic similarity, temporal proximity, and event dependence between events, including:

[0128] Let the event correlation be S event , the semantic similarity between events be S semantic , the temporal proximity between events be S time , the event dependence between events be S dependency , the weight of the semantic similarity be α, the weight of the temporal proximity be β, and the weight of the event dependence be γ, then:

[0129] S event = αSsemantic +βS time +γS dependency 。

[0130] Optionally, according to the event correlation degree, use the dynamic programming algorithm to adjust the priority and reminder time of the event, including:

[0131] Take the priority of the event as the state variable, set the completion time of the event and the resource usage as the benefit parameters, construct the state transition equation, and calculate the state variable corresponding to the maximum benefit; set the time at a certain duration from the deadline as the reminder time.

[0132] Based on the dynamic event correlation degree, it can automatically adjust the priority of the event as the event progresses, thereby adjusting the event management strategy. At the same time, combined with data analysis and prediction, it realizes intelligent reminder and optimization of the event.

[0133] Optionally, adopt an adaptive learning algorithm to adjust the event reminder strategy according to the user's feedback data, including:

[0134] Collect the event reminder strategy in historical events and the user's feedback data; the user's feedback data includes the event creation time, event completion time, and event priority; the event reminder strategy includes the reminder time and reminder method;

[0135] Construct an event reminder model based on the machine learning algorithm. The event reminder model extracts feature data from the user's feedback data, takes the reminder time and reminder method as the output of the event reminder model, trains the event reminder model, updates the parameters of the event reminder model, and obtains the target event reminder model;

[0136] Input the user's feedback data corresponding to the target event into the target event reminder model to obtain the event reminder strategy of the target event.

[0137] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for quantifying and modeling the correlation degree of dynamic events, characterized in that, Including: Structurally process the recorded text and represent it in vector form; Identify events and extract the time information of the events; Calculate the semantic similarity between events, and at the same time calculate the temporal proximity between events based on the time information of the events; Assign weights to events according to the dependency relationship between events to obtain the event dependency between events; Determine the event correlation degree according to the semantic similarity, temporal proximity and event dependency between events; Adjust the priority and reminder time of events using the dynamic programming algorithm according to the event correlation degree; Construct a Bayesian network based on the event stream, calculate the probability of completion of each node of the event, and predict the probability of event completion; Adopt an adaptive learning algorithm to adjust the event reminder strategy according to the user's feedback data.

2. The dynamic event relevance quantification modeling method according to claim 1, characterized in that Use natural language processing algorithms to structurally process the recorded text, including: Segment the recorded text, annotate it according to the part of speech, identify the entities in the recorded text, extract the key information from the recorded text, and organize it according to the preset structural format; Extract the mapping relationship between the text features and structural information of the key information, construct a machine learning model and perform model training, and use the machine learning model to structurally process the key information of the recorded text.

3. The dynamic event correlation degree quantification modeling method according to claim 1, wherein Use a pre-trained language model to represent the recorded text in vector form.

4. The dynamic event relevance quantification modeling method according to claim 1, characterized in that Identify events, including: Describe events using a triple structure; the triple structure includes time information, action information and object; Construct a deep learning model, including a BERT model, a bidirectional long short-term memory network and a conditional random field layer; Collect text data and manually annotate the text data related to events; Divide the annotated text data into a training set, a validation set and a test set; Load the pre-trained BERT model and convert the text data into a sequence of word vectors; The BERT model performs bidirectional context feature extraction on the input text to capture the semantic information in the text; Input the feature vectors output by the BERT model into the bidirectional long short-term memory network, and the bidirectional long short-term memory network extracts the semantic features of the text and outputs the hidden layer state; Use the output of the bidirectional long short-term memory network as the input of the conditional random field layer, and the conditional random field layer constrains and optimizes the prediction results of the bidirectional long short-term memory network to obtain the final event extraction result; Use the training set to train the deep learning model and adjust the parameters of the model by minimizing the loss function; Obtain the event recognition result using the trained deep learning model.

5. The dynamic event correlation degree quantification modeling method according to claim 1, wherein Calculate the semantic similarity between events, including: calculating the cosine similarity of the vectors corresponding to the events to obtain the semantic similarity between events.

6. The dynamic event correlation degree quantification modeling method according to claim 1, characterized in that Use an exponential decay function to calculate the temporal proximity between events.

7. The dynamic event correlation degree quantification modeling method according to claim 1, characterized in that Determine the event correlation degree according to the semantic similarity, temporal proximity and event dependency between events, including: Let the event correlation degree be S event , the semantic similarity between events be S semantic , the time proximity between events be S time , the event dependence degree between events be S dependency , the weight of semantic similarity be α, the weight of time proximity be β, and the weight of event dependence degree be γ, then: S event = αS semantic + βS time + γS dependency 。 8. The dynamic event correlation degree quantization modeling method according to claim 1, characterized in that According to the event correlation degree, use the dynamic programming algorithm to adjust the priority and reminder time of events, including: Take the priority of the event as a state variable, set the completion time of the event and the resource usage as benefit parameters, construct a state transition equation, and calculate the state variable corresponding to the maximum benefit; set the time at a duration from the deadline as the reminder time.

9. The dynamic event correlation degree quantification modeling method according to claim 1, wherein Adopt an adaptive learning algorithm to adjust the event reminder strategy according to the user's feedback data, including: Collect the event reminder strategies and the user's feedback data in historical events; the user's feedback data includes the event creation time, event completion time, and event priority; the event reminder strategy includes the reminder time and reminder method; Construct an event reminder model based on a machine learning algorithm. The event reminder model extracts feature data from the user's feedback data, takes the reminder time and reminder method as the output of the event reminder model, trains the event reminder model, updates the parameters of the event reminder model, and obtains the target event reminder model; Input the feedback data of the user corresponding to the target event into the target event reminder model to obtain the event reminder strategy of the target event.

10. A dynamic event correlation degree quantification and modeling system, characterized in that It includes a text processing unit, an identification and extraction unit, a first analysis and processing unit, a second analysis and processing unit, a data processing unit, a first adjustment unit, a prediction unit, and a second adjustment unit; The text processing unit is used to perform structured processing on the recorded text and perform vectorization representation; The identification and extraction unit is used to identify the event and extract the time information of the event; The first analysis and processing unit is used to calculate the semantic similarity between events and calculate the time proximity between events according to the time information of the events; The second analysis and processing unit is used to assign weights to the events according to the dependency relationship between events to obtain the event dependency between events; The data processing unit is used to determine the event correlation according to the semantic similarity, time proximity, and event dependency between events; The first adjustment unit is used to adjust the priority and reminder time of the event using a dynamic programming algorithm according to the event correlation; The prediction unit is used to construct a Bayesian network based on the event stream, calculate the probability of completion of each node of the event, and predict the probability of event completion; The second adjustment unit is used to adopt an adaptive learning algorithm to adjust the event reminder strategy according to the user's feedback data.

Images