A data classification framework based on causal feature selection

The performance of the classification model is related to the quality of the data, so it is necessary to effectively preprocess the data to improve the classification effect. Because traditional classification methods do not consider the causality between attributes, it is difficult to eliminate features that are not related to the target variable. At the same time, the existing methods for measuring the distance between samples are mostly based on the Euclidean distance, which is difficult to adapt to the characteristics of the samples processed by causal inference. In order to solve the above problems and achieve stable classification of arbitrary datasets, a classification framework was proposed. The framework uses the causal inference algorithm to construct a causal map, screens out the features with strong correlation with the target variable, and furthermore, evaluates the similarity between the sample data based on the Pearson correlation coefficient to find similar neighbors, and calculates the predicted value with the similarity weight and category value to predict the target variable more accurately. The experimental results show that the proposed classification framework is superior to the traditional classification model in terms of performance.