The invention is applicable to the technical field of
machine learning, and provides a multivariate discrete
feature selection method, device, apparatus, and storage medium. The method comprises: initializing a particle swarm according to a tangent point corresponding to each feature in a target
data set, obtaining the
particle position of each particle, the target
data set is discretized according to the
particle position, the corresponding discrete
data set is obtained, in accordance with discrete data set, the fitness of each particle is calculated by the fitness formula, so as to find outthe optimal position of the
population of the particle swarm and the optimal position of the individual through which each particle passes, when the stop condition is met, optimal position of output
population, otherwise, the
particle position of each particle is updated according to the optimal position of the
population and the optimal position of the individual, and the operation of data
discretization and optimization is continued, so that fewer features are selected, the effect of eliminating redundant features and irrelevant features is improved, and the accuracy of classification learning
algorithm is improved.