Machine Learning Model for Data Credibility Assessment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies fail to accurately determine the credibility of internet data for recipients, as they do not consider the reliability of the provider's information, leading to inappropriate credibility assessments based on the recipient's attributes and expertise.
Innovation Solution
An information processing system that uses a trained machine learning model to assess credibility by generating reliability information based on a reliability list and a credibility determination model, taking into account the recipient's specific reliability list and evaluation tendencies, and updates the models based on feedback to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing technologies are used to determine data credibility, then the process is simple and quick, but the accuracy of credibility assessment is low because recipient-specific reliability factors are not considered
Solution Approach 1:
The credibility assessment system is segmented into multiple independent modules: a reliability list management module that stores recipient-specific reliable generator information, a machine learning model processing module that analyzes property information, and a credibility determination module that integrates both reliability information and ML outputs. This segmentation allows the system to incorporate complex recipient-specific factors without overwhelming the overall architecture, improving accuracy while managing complexity through modular design.
Solution Approach 2:
The system performs preliminary actions by pre-storing reliability information about information generators in a reliability list specific to each recipient, and by pre-training machine learning models with historical data before actual credibility assessment. This preliminary preparation enables the system to quickly and accurately assess credibility during actual use without having to process all factors from scratch, thereby improving assessment accuracy while maintaining operational efficiency.
2Measurement precision
If a trained machine learning model with reliability information is used, then credibility determination accuracy is improved, but the processing time and computational resources increase
Solution Approach 1:
The system applies partial action by selectively using different assessment approaches based on the situation. For routine assessments, it may rely more on the pre-trained machine learning model's predictions. For cases requiring higher accuracy or when dealing with unfamiliar information generators, it incorporates the detailed reliability list information. This selective application of computational resources improves accuracy when needed while reducing processing time for routine cases.
Solution Approach 2:
The machine learning model is pre-trained offline using historical credibility data and recipient profiles, performing computationally intensive work before actual credibility assessments. During runtime, the pre-trained model provides quick predictions that are then refined by incorporating reliability list information, rather than performing full training and analysis during each assessment, thus reducing processing time while maintaining high accuracy.
3Adaptability or versatility
If reliability information from a reliability list is generated and integrated, then personalized credibility assessment is achieved, but the system complexity and data management requirements increase
Solution Approach 1:
The reliability list structure is designed as a universal data framework that can store various types of information about different information generators (news outlets, social media accounts, websites, etc.) in a standardized format. This universal structure serves multiple functions: storing reliability ratings, tracking recipient preferences, recording historical interactions, and supporting different types of credibility assessments. This multi-functionality enables personalized assessment across diverse scenarios without requiring separate systems for each case, managing complexity through standardized universal data structures.
Data Source
AI summary
A non-transitory computer-readable recording medium stores an information processing program for causing a computer to execute a process. The computer includes a trained first machine learning model that receives, as an input, property information which indicates a probability of data and is assigned reliability information which indicates reliability of the property information, and outputs credibility of the data. The process includes: acquiring first property information that indicates a probability of predetermined data received by a recipient; generating first reliability information that indicates reliability of the first property information, based on a reliability list in which a generator reliable for the recipient among generators of the first property information is registered; and inputting, to the first machine learning model, the first property information that is assigned the first reliability information, determining credibility of the predetermined data, and presenting a result.