A mobile office application-oriented intelligent dynamic integration method and system

Abstract

This invention discloses an intelligent dynamic integration method and system for mobile office applications, addressing issues such as fragmented mobile office applications, inefficient information integration, and low recommendation matching accuracy. The method includes collecting user attributes and application usage data; establishing a user attribute mapping layer with a hash table structure and assigning initial weights to applications; constructing a neural network recommendation model with a multilayer perceptron structure; obtaining real-time application weights after supervised learning training and inputting updated data; sorting and grouping applications according to their weights; and then accurately pushing applications to users through the mapping layer. The system includes corresponding acquisition, mapping, modeling, classification, and push units, and also provides electronic devices and storage media for implementing the method. This invention achieves personalized and accurate matching between applications and user needs, improves the real-time performance and stability of recommendations, adapts to dynamic changes in enterprise organizational structure and user needs, and significantly improves mobile office efficiency.

Description

Technical Field

[0001] This invention relates to the field of intelligent recommendation technology, specifically to an intelligent dynamic integration method and system for mobile office applications. Background Technology

[0002] In today's global digital wave, the rapid popularization of 5G networks and the profound innovation of cloud computing technology have injected strong momentum into the booming development of the mobile internet. Mobile office has become an indispensable and important mode of modern enterprise operation. For enterprises, mobile office is not only a tool to break through the limitations of time and space and improve the flexibility of office work, but also a core engine to promote enterprise digital transformation and enhance market competitiveness. To meet the diversified business needs of enterprises, various mobile office applications have sprung up like mushrooms after rain. In the field of office approval, enterprises have developed dedicated mobile approval applications to realize the online processing of processes such as procurement applications and expense reimbursements; in terms of information access, mobile access applications covering official documents, announcements, production and operation data, etc., help employees obtain key information anytime, anywhere; in terms of communication and collaboration, mobile applications such as instant messaging and video conferencing break down communication barriers between team members. The original intention of these applications is to build a convenient and efficient office environment for employees through digital means and accelerate the operation of enterprise business processes.

[0003] Enterprises possess a large number of scattered mobile office applications, lacking a unified and efficient integrated display solution. Existing integration methods struggle to meet the diverse application display needs, failing to offer flexible presentation through rich formats, and also hindering application grouping management and dynamic sorting based on usage habits. This makes it difficult for employees to quickly locate and use the applications they need, impacting work efficiency.

[0004] Long-term practice has revealed numerous technical challenges in customizing mobile portals for enterprise mobile office scenarios. On one hand, a large number of mobile office applications within enterprises are scattered, lacking a unified and efficient integration and display solution. Existing integration methods struggle to meet diverse needs, cannot flexibly present applications, and are unable to achieve group management and dynamic sorting, making it difficult for employees to quickly locate and use applications, thus impacting work efficiency. On the other hand, announcements, official documents, and other information are scattered across different systems, making integration on mobile devices difficult and displaying information unintuitively. For example, announcements cannot distinguish between read and unread status, official document formats are poorly adapted, and data display lacks visualization, making it inconvenient for employees to access information and hindering the full realization of its value. Furthermore, employee to-do lists are scattered, lacking unified integrated management and intelligent reminders, making it difficult to achieve socialized office work and reducing the timeliness of task processing and the efficiency of collaborative work. Finally, existing information portals struggle to meet the needs of multi-tenant models, cannot customize content, and have incomplete information publishing, approval, and display functions, lacking interactive features, resulting in low information dissemination efficiency and failing to meet the diverse information access and communication needs of employees. Summary of the Invention

[0005] This invention aims to overcome the shortcomings of existing technologies and proposes an intelligent dynamic integration method for mobile office applications, comprising the following steps:

[0006] Step S1: Obtain user attribute data and application usage data. The user attribute data includes at least basic user information, department, position, and responsibilities. The application usage data includes at least usage duration and preference data. The user attribute data and application usage data are collected periodically, or when the user attribute data or application usage data changes.

[0007] Step S2: Establish a user attribute mapping layer between the user attribute data and the application, and classify the application according to different user attribute data and assign initial weights. The user attribute mapping layer uses a hash table data structure for storage, with the feature combination of user attribute data as the key and the corresponding application and initial weight as the value.

[0008] Step S3: Establish a recommendation model based on a neural network, taking the application usage data and the user attribute data as inputs, and the initial weights of the application as outputs, and train the recommendation model in a supervised learning mode. When training the recommendation model, the stochastic gradient descent optimization algorithm is used to adjust the model parameters, and the mean squared error loss function is selected as the loss function.

[0009] The neural network in the recommendation model based on neural networks adopts a multilayer perceptron structure, which includes an input layer, a hidden layer, and an output layer, with at least two hidden layers.

[0010] Step S4: Input the updated user attribute data and application usage data into the trained recommendation model to obtain the real-time weights of the applications; sort and group the applications according to the real-time weights. The application usage data is obtained through the terminal device used by the user.

[0011] Step S5: The sorted and grouped applications are pushed to different users through the user attribute mapping layer.

[0012] Based on the above-described intelligent dynamic integration method for mobile office applications, this invention provides a system according to the above-described intelligent dynamic integration method for mobile office applications, the system comprising: The acquisition unit is used to acquire user attribute data and application usage data. The user attribute data includes at least basic user information, department, position, and responsibilities. The application usage data includes at least usage duration and preference data. The mapping unit is used to establish a user attribute mapping layer between the user attribute data and the application, and to classify the application and assign initial weights according to different user attribute data. The model unit is used to build a recommendation model based on a neural network. It takes the application usage data and the user attribute data as input and the initial weights of the application as output, and trains the recommendation model in a supervised learning mode. The classification unit is used to input the updated user attribute data and application usage data into the trained recommendation model to obtain the real-time weights of the applications; and to sort and group the applications according to the real-time weights. The push unit is used to push the sorted and grouped applications to different users through the user attribute mapping layer.

[0013] The acquisition unit includes a data acquisition module, which is used to periodically collect the user attribute data and the application usage data, or to collect the data when the user attribute data or the application usage data changes; the application usage data is collected through the user terminal device.

[0014] The present invention provides an electronic device, at least one processor, and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to execute the above-described intelligent dynamic integration method for mobile office applications.

[0015] The present invention provides a machine-readable storage medium storing instructions that cause a machine to execute the intelligent dynamic integration method for mobile office applications as described above.

[0016] Compared with the prior art, the beneficial effects of the present invention include: (1) By establishing a user attribute mapping layer, the office applications are classified and their weights are dynamically adjusted based on user basic information, department, position, responsibilities and other attributes. This enables the application to accurately match the personalized needs of different users in the enterprise, thus solving the problem of scattered mobile office applications and low matching degree with user needs.

[0017] (2) The neural network recommendation model constructed by the multilayer perceptron can deeply mine the feature association between user attributes and application usage data through supervised learning training, making the application weight calculation more in line with the actual office usage habits of users, and significantly improving the recommendation accuracy compared with traditional methods.

[0018] (3) By adopting a hybrid data collection method of real-time and periodic data collection, the dynamic changes of user attributes and work habits can be captured in a timely manner; the real-time weights of the model output are smoothed and then sorted and grouped, which not only ensures the real-time update of the recommendation results, but also avoids the instability of recommendations caused by drastic fluctuations in weights.

[0019] (4) Introduce a user feedback mechanism to apply user feedback to the user attribute mapping layer and recommendation model, continuously adjust the mapping rules and model parameters, so that the system can adapt to long-term changes in enterprise organizational structure and user office needs and maintain long-term applicability. Attached Figure Description

[0020] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0021] Figure 1 This is a flowchart illustrating an intelligent dynamic integration method for mobile office applications according to the present invention.

[0022] Figure 2 This is a schematic diagram of the structure of an intelligent dynamic integration system for mobile office applications according to the present invention. Detailed Implementation

[0023] Example 1 The intelligent dynamic integration method for mobile office applications described in this invention has the following overall process: Figure 1 As shown, the specific steps are as follows: Step S1: Obtain user attribute data and application usage data. The user attribute data includes at least basic user information, department, position, and responsibilities. The application usage data includes at least usage duration and preference data. To trigger data collection only when data changes, avoiding meaningless periodic full updates, and reducing server storage and computing resource consumption, while also considering periodic data collection for analyzing user behavior trends, in a more preferred embodiment of the invention, user attribute data and application usage data are collected periodically, or when these data change. Changes in user habits, job responsibilities, or departmental adjustments may lead to changes in needs; for example, an employee promoted to use more management applications. Real-time data collection allows for timely updates to the recommendation model, avoiding recommendation lag. Example: If an employee is promoted from development engineer to technical manager, the system automatically increases the recommendation weight of project management applications by collecting job change data.

[0024] Step S2: Establish a user attribute mapping layer between the user attribute data and the application, and classify the application according to different user attribute data and assign initial weights. The system collects basic information such as name, age, and contact information from user registration / login forms. Structured data including departments, positions, and responsibilities is synchronized from internal enterprise systems, such as OA and HR systems. User behavior is monitored and recorded within the application, including application duration and frequency of function module clicks. Real-time data on usage duration, access paths, and function preferences is collected through a log system, such as frequent use of reporting functions in office software. A "user attribute-application" association matrix is ​​constructed; for example, a user with the department "Technology Department" and position "Development Engineer" is mapped to programming tools and code management applications. Users whose responsibilities include "data reporting" are mapped to report generation and data visualization applications. A graph database, Neo4j, is used to store the mapping relationships, with nodes representing "user attributes" and "applications," and edges representing association weights, initially set to default values ​​of 0-1. Weights are preset based on a rule engine; for example, the initial weight for a "Technology Department Development Engineer" is set to 0.8 for programming tools and 0.2 for marketing applications. Administrators can manually adjust weights to suit enterprise needs.

[0025] Step S3: Establish a recommendation model based on a neural network, taking the application usage data and the user attribute data as inputs, and the initial weights of the application as outputs, and train the recommendation model in a supervised learning mode. In the input layer of the model architecture design, user attribute data, such as department and position, are encoded as one-hot (one-bit valid code) vectors or embedding vectors. For example, "Technical Department" is encoded as [0,1,0,0], if the company has 4 departments. Application usage data, such as usage duration and preference labels, are normalized into numerical features, such as standardizing usage duration to the [0,1] interval.

[0026] The hidden layers in the model architecture design adopt a multilayer perceptron (MLP) structure, which uses activation functions such as ReLU to capture the nonlinear relationships between features. The number of layers and neurons is adjusted according to the data scale, such as 2-3 layers, with 128-256 neurons in each layer.

[0027] In the model architecture design, the output layer has the number of applications as its output dimension. Each value represents the initial weight of the corresponding application and is mapped to the [0,1] interval through Softmax (soft maximum function) or a linear layer.

[0028] During model training, the correspondence between user attributes, usage data, and actual application weights in historical data is collected as labeled samples. For example, the actual usage weight of application B corresponding to user A's attributes is 0.7. The loss function and optimizer use mean squared error (MSE) to measure the deviation between predicted and actual weights, or cross-entropy loss. The optimizer uses algorithms such as Adam and SGD to iteratively update model parameters and reduce the loss value. User attribute data and application usage data are loaded and merged using pandas. The numerical features are standardized using sklearn's ColumnTransformer, and categorical features are one-hot encoded.

[0029] Step S4: Input the updated user attribute data and application usage data into the trained recommendation model to obtain the real-time weights of the applications; sort and group the applications according to the real-time weights. Step S5: The sorted and grouped applications are pushed to different users through the user attribute mapping layer.

[0030] To ensure the real-time performance of recommendation services under high-concurrency scenarios and avoid response timeouts due to data query blocking, the user attribute mapping layer in this invention employs a hash table data structure for storage. The user attribute data feature combinations serve as keys, while the corresponding application and initial weight are used as values. The hash function maps user attribute feature combinations, such as "department = technical department + position = development engineer," to unique hash values, enabling millisecond-level queries of application categories and weights. This represents at least a 10-fold improvement in efficiency compared to traversal matching methods such as list or tree structures. The hash table supports insertion and deletion operations with minimal time complexity, facilitating administrators to add or modify user attribute-application mapping rules.

[0031] To ensure efficient and stable gradient calculation, suitable for enterprise-level applications where user behavior data changes dynamically and recommendation results require rapid response, this invention provides stable and high-precision recommendation services while maintaining low computational costs. In a more preferred embodiment, when training the recommendation model, the stochastic gradient descent (SGD) algorithm is used to adjust the model parameters, and the mean squared error loss function is chosen as the loss function. The mean squared error loss function (MSE) is defined as the average of the sum of squares of the differences between the predicted and true values. When combined with the stochastic gradient descent (SGD) algorithm, the computational complexity of each parameter update is only O(1), suitable for the rapid iteration of real-time recommendation systems. The stochastic gradient descent algorithm randomly selects a single sample or a small batch of samples to calculate the gradient, and then iteratively updates the model parameters to minimize the loss function. Compared to calculating the gradient using all samples, SGD significantly reduces the computational cost of each iteration, accelerating model training. By setting a dynamic learning rate, the rapid convergence in the early stages of training and the fine-tuning in the later stages can be balanced.

[0032] In a more preferred embodiment of this invention, the recommendation model employs DIN (Deep Interest Network), DIEN (Deep Interest Evolution Network), or SASRec (Self-Attentive Sequential Recommendation), which can capture long-distance dependencies in behavioral sequences using a self-attention mechanism, avoiding the temporal bias of RNNs. Simultaneously optimizing multiple objectives, such as click-through rate (CTR) and dwell time, avoids single-objective bias. ESMM (Entire Space Multi-Task Model) is preferred, modeling CTR and other metrics as multiple tasks, resolving sample selection bias in the CTR task through sample space sharing. Alternatively, MMoE (Multi-gate Mixture-of-Experts) can be used, processing different sub-tasks through multiple expert networks and dynamically allocating weights using a gating mechanism, suitable for scenarios where objectives conflict.

[0033] For lightweight models, using Factorization Machine (FM) or Field-weighted Factorization Machine (FwFM) can efficiently learn low-order feature interactions with a complexity of O(n) (where n is the number of features), making them suitable as a base model or for use in DeepFM combined with deep learning. In scenarios with sufficient feature engineering, linear models are fast and highly interpretable, and the baseline model LinearSVM is used.

[0034] To directly monitor application runtime from terminal devices such as mobile phones, computers, and tablets, avoiding delays caused by server relays (e.g., network transmission time, log aggregation processing latency), and achieving second-level data collection, this invention, in a more preferred embodiment, obtains application usage data directly from the user's terminal device. Application usage data is obtained through terminal operating system APIs, such as Android's UsageStats, iOS's ScreenTime, and Windows' AppUsage, accurate to the millisecond level for usage duration and background / foreground state switching records, which is more precise than server logs. This allows for a better construction of a comprehensive behavioral profile. For example, if a user frequently uses instant messaging applications on a mobile device and frequently uses document collaboration applications on a computer, the recommendation system can dynamically adjust recommendation priorities based on these different devices.

[0035] To automatically learn higher-order interactions between features, this method is suitable for handling massive amounts of heterogeneous features. In a more preferred embodiment of this invention, the neural network in the recommendation model employs a multilayer perceptron (MLP) structure, comprising an input layer, hidden layers, and an output layer, with at least two hidden layers. While a single-layer perceptron can only handle linearly separable problems, such as logistic regression, the MLP introduces nonlinear transformations through activation functions in the hidden layers, such as ReLU and Sigmoid, enabling it to fit the complex mapping relationship between user behavior and recommendation results. For example, the correlation between features such as user age, spending power, and historical click behavior and application preferences is often nonlinear. For instance, a 25-year-old user's preference for gaming applications does not decrease linearly with age but exhibits an inflection point. The MLP captures these "nonlinear inflection points" through multilayer computation.

[0036] Example 2 Based on the intelligent dynamic integration method for mobile office applications described in Embodiment 1, a system for intelligent dynamic integration of mobile office applications is provided, the structure of which is as follows: Figure 2 As shown, the system includes, The acquisition unit is used to acquire user attribute data and application usage data. The user attribute data includes at least basic user information, department, position, and responsibilities. The application usage data includes at least usage duration and preference data. The mapping unit is used to establish a user attribute mapping layer between the user attribute data and the application, and to perform mapping on the application according to different user attribute data; The model unit is used to build a recommendation model based on a neural network. It takes the application usage data and the user attribute data as input and the initial weights of the application as output, and trains the recommendation model in a supervised learning mode. The classification unit is used to input the updated user attribute data and application usage data into the trained recommendation model to obtain the real-time weights of the applications; and to sort and group the applications according to the real-time weights. The push unit is used to push sorted and grouped applications to different users through the user attribute mapping layer.

[0037] The system comprises an acquisition unit, a mapping unit, a modeling unit, a classification unit, and a push unit. The acquisition unit obtains user attribute data such as basic user information, department, position, and responsibilities, as well as application usage data such as usage time and preferences. The mapping unit establishes a mapping layer between user attribute data and applications, classifies applications based on user attributes, and assigns initial weights. The modeling unit constructs a neural network-based recommendation model, using application usage data and user attribute data as input and initial weights as output for supervised learning training. The classification unit inputs updated data into the trained model to obtain real-time weights and sorts and groups applications accordingly. The push unit accurately pushes the processed applications to different users via the mapping layer. Through a hybrid recommendation model combining data-driven and rule-assisted approaches, the system achieves a balance between recommendation accuracy and business adaptability while ensuring project feasibility, making it particularly suitable for enterprise-level application scenarios with complex organizational structures and dynamically changing user needs.

[0038] To capture user attributes in real time, such as department, position, or dynamic changes in usage behavior, such as duration and preferences, and to avoid recommendation failure due to data lag (e.g., timely updates to recommendation strategies after a user changes jobs), terminal device data collection ensures data timeliness and completeness. Simultaneously, the collected device information can be combined with application data to optimize recommendations, such as adjusting application weights based on device performance, thereby improving user experience and recommendation accuracy. In a more preferred embodiment of the invention, the acquisition unit includes a data collection module, which is used to periodically collect the user attribute data and the application usage data, or to collect data when the user attribute data or the application usage data changes; the application usage data is collected through the user's terminal device.

[0039] The present invention provides an electronic device, at least one processor, and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to execute the above-described intelligent dynamic integration method for mobile office applications.

[0040] The present invention provides a machine-readable storage medium storing instructions that cause a machine to execute the intelligent dynamic integration method for mobile office applications as described above.

[0041] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, because according to the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0042] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0043] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical or other forms.

[0044] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0045] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0046] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, mobile terminal, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0047] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent dynamic integration for mobile office application, characterized in that, Includes the following steps: S1. Obtain user attribute data and application usage data, wherein the user attribute data includes basic user information, department, position, and responsibilities; and the application usage data includes usage duration and preference data. S2. Establish a user attribute mapping layer between the user attribute data mentioned in S1 and the application, and classify the application according to different user attribute data and assign initial weights; S3. Establish a recommendation model based on a neural network, taking the application usage data and the user attribute data as inputs, and the initial weights of the application as outputs, and train the recommendation model in a supervised learning mode; S4. Input the updated user attribute data and application usage data into the trained recommendation model to obtain the real-time weights of the application; Applications are sorted and grouped based on real-time weights; S5. The sorted and grouped applications are pushed to different users through the user attribute mapping layer.

2. The method of claim 1, wherein, In step S1, a hybrid acquisition method is used to obtain data, that is, the user attribute data and application usage data are periodically acquired, or acquisition is triggered when the user attribute data or application usage data changes.

3. The method of claim 1, wherein the mobile office application oriented intelligent dynamic integration method is characterized by, The user attribute mapping layer in step S2 uses a hash table data structure for storage, with the feature combination of user attribute data as the key and the corresponding application and initial weight as the value.

4. The intelligent dynamic integration method for mobile office applications according to claim 1, characterized in that, In step S3, when training the recommendation model, the stochastic gradient descent optimization algorithm is used to adjust the model parameters, and the mean squared error loss function is selected as the loss function.

5. The intelligent dynamic integration method for mobile office applications according to claim 1, characterized in that, When collecting application usage data in step S1, the application usage data is obtained through the operating system interface of the user's terminal device.

6. The intelligent dynamic integration method for mobile office applications according to claim 1, characterized in that, The neural network in the recommendation model based on neural networks in step S3 adopts a multilayer perceptron structure, which includes an input layer, a hidden layer and an output layer, with at least two hidden layers.

7. The intelligent dynamic integration method for mobile office applications according to claim 6, characterized in that, The input layer encodes the classification features of the user attribute data input to the recommendation model into a one-bit valid encoding vector or embedding vector, and normalizes the application data into numerical features in the [0,1] interval.

8. The intelligent dynamic integration method for mobile office applications according to claim 6, characterized in that, The output layer is mapped to the [0,1] interval through a soft maximum function or a linear layer. The output dimension is the number of applications, and each value represents the initial weight of the corresponding application.

9. A system for an intelligent dynamic integration method for mobile office applications according to any one of claims 1-8, characterized in that, The system includes an acquisition unit, a mapping unit, a model unit, a classification unit, and a push unit; The acquisition unit is used to acquire user attribute data and application usage data. The user attribute data includes basic user information, department, position, and responsibilities. The application usage data includes usage duration and preference data; The mapping unit is used to establish a user attribute mapping layer between the user attribute data and the application, and to classify the application and assign initial weights according to different user attribute data. The model unit is used to establish a recommendation model based on a neural network. It takes the application usage data and the user attribute data as input, the initial weights of the application as output, and trains the recommendation model in a supervised learning mode. The classification unit is used to input the updated user attribute data and application usage data into the trained recommendation model to obtain the real-time weight of the application, and to sort and group the applications according to the real-time weight. The push unit is used to push the sorted and grouped applications to different users through the user attribute mapping layer.

10. An electronic device, characterized in that, It includes at least one processor and a memory communicatively connected to the at least one processor; the memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enable the at least one processor to perform the intelligent dynamic integration method for mobile office applications as described in any one of claims 1-8.

Images