Hybrid Neural Network for Application Preloading Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for preloading applications on intelligent terminals are inefficient, leading to increased storage space usage and power consumption, affecting the terminal's fluency and user experience, as they do not accurately predict which applications will be launched next.
Innovation Solution
A method using a hybrid network model, combining a recurrent-neural-network submodel and a fully connected network submodel, to predict the next application to be launched based on current usage information and historical data, optimizing resource allocation and preloading processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If applications are preloaded in advance to improve loading speed, then application loading time is reduced, but storage space consumption and power consumption increase
Solution Approach 1:
The system performs preliminary actions by preloading applications in advance based on prediction algorithms. The hybrid network model predicts which applications will be launched next and preloads them into memory, so that when the user actually launches the application, it is already loaded and can start immediately, reducing loading time while avoiding unnecessary preloading of applications that won't be used
Solution Approach 2:
The system changes the parameter of preloading strategy from static (preloading fixed numbers of applications) to dynamic (predicting which applications to preload based on historical usage data, current context, and machine learning models). This allows the system to adaptively adjust which applications are preloaded, optimizing the balance between loading speed and power consumption
2Ease of operation
If multiple applications are preloaded to improve user experience, then application availability is improved, but storage space occupation increases
Solution Approach 1:
The system performs preliminary actions by preloading applications in advance based on prediction algorithms. The hybrid network model predicts which applications will be launched next and preloads them into memory, so that when the user actually launches the application, it is already loaded and can start immediately, reducing loading time while avoiding unnecessary preloading of applications that won't be used
Solution Approach 2:
The system applies partial action by selectively preloading only the most likely candidate applications rather than preloading all applications. The hybrid network model ranks applications by prediction probability and preloads only the top candidates, avoiding the excessive consumption of storage space while still improving application availability for the most important cases
3Device complexity
If a simple preloading mechanism is used, then system complexity is reduced, but prediction accuracy of next application decreases
Solution Approach 1:
The system merges multiple machine learning models (recurrent neural network for sequence prediction and fully connected network for feature processing) into a hybrid network model. This combination leverages the strengths of different model types to improve prediction accuracy while maintaining a unified, manageable architecture that doesn't excessively increase system complexity
Solution Approach 2:
The prediction mechanism is segmented into distinct functional components: data collection module, hybrid network model module, and preloading execution module. This segmentation allows each component to be optimized independently and simplifies the overall system architecture, making the complex prediction process more manageable and maintainable
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method for preloading an application, a storage medium, and a terminal are provided. The method includes the following. Current usage information of a terminal is acquired in response to detecting that an event of prediction of preloading an application is triggered (101). The current usage information is input to a model used for application prediction (102). The model used for application prediction includes a hybrid network model generated according to a usage rule of applications of the terminal corresponding to historical status feature information (103). At least one target application to be launched is predicted according to an output result of the model used for application prediction (104). The at least one target application is preloaded.