Presentation Information
[P04-581]Research on the Construction and Application of Artificial Intelligence-Based Agent Tools for Health Economic
○xueling wang1, YueYao Chen1 (1. Ninth People's Hospital Shanghai Jiao Tong University School of Medicine (China))
Keywords:
Artificial intelligence,Economic accounting,Intelligent agent,biological sample
[Purpose]To develop an intelligent agent tool incorporating a novel accounting method, addressing the inefficiency in cost accounting for biobanks in existing technologies and achieving the goal of improving cost accounting efficiency.
[Method]Obtain initial cost data and generate cost vectors based on the initial data; generate input vectors according to the cost vectors, and analyze the input vectors via a multi-task neural network to obtain the first cost predicted values, including at least direct costs and indirect costs. Analyze at least the first cost predicted values through a target residual network to obtain a weight matrix, and allocate indirect costs via the weight matrix to obtain the second cost predicted values for each biological sample. Calculate the total predicted cost of each biological sample based on the first and second cost predicted values. The direct and indirect costs of each biological sample are predicted separately through the multi-task neural network and the target residual network, with indirect cost allocation implemented. Through closed-loop model learning and training, error evaluation results trigger an incremental learning strategy. Based on historical final accounts data and future plans, Monte Carlo simulations are conducted in a digital twin sandbox to generate budget distributions, which are then compared with actual final account distributions. Reinforcement learning is used to optimize resource scheduling in the simulation. When the budget-final account discrepancy is excessively large, model incremental learning is triggered to form a closed loop. In the cost audit process, audit query requests are received through endpoints, and relevant cost reports are retrieved (possibly from local databases or off-chain storage). The corresponding on-chain evidence transactions (including root hash, metadata, signature, block height, transaction ID tx_id) are retrieved from the blockchain. The consistency between the root hash recorded on the chain and that calculated from the locally retrieved report is verified, providing a complete and verifiable audit trail.
[Results]Focusing on the core pain point of low efficiency in biobank cost accounting, this study constructs an AI-powered health economic accounting agent tool. Following the development path of "data-model-training-application-validation", the model is optimized via closed-loop learning and training. Combined with the cost audit process and adaptive environment configuration, a complete cost accounting system is formed. Verification results show that the tool, with the mechanism of "dual-model collaborative prediction + precise allocation", is expected to effectively solve the pain points of traditional accounting, improve the efficiency and accuracy of biological sample cost accounting, and provide feasibility for the application of artificial intelligence in the health economy of biological samples.
[Consideration]Data quality, model interpretability, privacy security and ethical norms should be fully considered to ensure the tool’s stability, reliability and compliance in health economic application.
[Conclusion]The artificial intelligence-based agent tool for health economic accounting of biological samples effectively solves the problem of low efficiency in cost accounting of biobanks in existing technologies.
[Method]Obtain initial cost data and generate cost vectors based on the initial data; generate input vectors according to the cost vectors, and analyze the input vectors via a multi-task neural network to obtain the first cost predicted values, including at least direct costs and indirect costs. Analyze at least the first cost predicted values through a target residual network to obtain a weight matrix, and allocate indirect costs via the weight matrix to obtain the second cost predicted values for each biological sample. Calculate the total predicted cost of each biological sample based on the first and second cost predicted values. The direct and indirect costs of each biological sample are predicted separately through the multi-task neural network and the target residual network, with indirect cost allocation implemented. Through closed-loop model learning and training, error evaluation results trigger an incremental learning strategy. Based on historical final accounts data and future plans, Monte Carlo simulations are conducted in a digital twin sandbox to generate budget distributions, which are then compared with actual final account distributions. Reinforcement learning is used to optimize resource scheduling in the simulation. When the budget-final account discrepancy is excessively large, model incremental learning is triggered to form a closed loop. In the cost audit process, audit query requests are received through endpoints, and relevant cost reports are retrieved (possibly from local databases or off-chain storage). The corresponding on-chain evidence transactions (including root hash, metadata, signature, block height, transaction ID tx_id) are retrieved from the blockchain. The consistency between the root hash recorded on the chain and that calculated from the locally retrieved report is verified, providing a complete and verifiable audit trail.
[Results]Focusing on the core pain point of low efficiency in biobank cost accounting, this study constructs an AI-powered health economic accounting agent tool. Following the development path of "data-model-training-application-validation", the model is optimized via closed-loop learning and training. Combined with the cost audit process and adaptive environment configuration, a complete cost accounting system is formed. Verification results show that the tool, with the mechanism of "dual-model collaborative prediction + precise allocation", is expected to effectively solve the pain points of traditional accounting, improve the efficiency and accuracy of biological sample cost accounting, and provide feasibility for the application of artificial intelligence in the health economy of biological samples.
[Consideration]Data quality, model interpretability, privacy security and ethical norms should be fully considered to ensure the tool’s stability, reliability and compliance in health economic application.
[Conclusion]The artificial intelligence-based agent tool for health economic accounting of biological samples effectively solves the problem of low efficiency in cost accounting of biobanks in existing technologies.
Comment
To browse or post comments, you must log in.Log in