It is important to note here, otherwise you will get wrong grouping results. 2. Engineering the calculation process of the causal inference model to improve reusability and shorten the development cycle Different models use basically the same feature variables.
Common feature variables can be fixed and automatically collected, enriching the feature variable library, which is convenient for improving the ws number list reusability of the model, shortening the development cycle, and efficiently giving strategy suggestions.
3. Iteratively optimize the logistic regression model, calculate the probability P and the weight coefficient w The propensity weighted score P is calculated by the commonly used logistic regression algorithm, and the categorical variables are one-hot encoded, and the matching weighted results are more uniform 1) Observe the significance of variables.