The Causal Inference of Bayesian Network Based on Improved IDA Algorithm

At present, the IDA algorithm is a commonly used method for estimating causal effects between variables from observational data. In the IDA algorithm, the PC algorithm is first used to estimate the directed acyclic graph (DAG) of the observational data, and then the causal effect values are calculated by combining this with the lower bound of the causal effects between variables, where the accuracy of estimating the directed acyclic graph directly affects the accuracy of the IDA algorithm's causal effect estimation. This paper improves the IDA algorithm based on the PCboot and Two-Stage methods, proposing the PCboot-IDA and TS-IDA algorithms. Research shows that the TS-IDA algorithm significantly outperforms the PCboot-IDA and IDA algorithms. Research shows that the new method improves the performance of IDA algorithm on both simulated and the T cell data set.