Abstract: In order to solve the problem of the reflection of corporate customer value, we proposed an improved model of TFA customer segmentation, which took customer development space T, purchase frequency F, and average purchase amount A as indicators to fully reflect the customer value and development space. Firstly, the K-means clustering algorithm was improved by introducing local density value ρ and information entropy H to optimize the traditional K-means clustering method in the initial clustering center selection problem. Secondly, by building a machine learning framework, clustering experiments were carried out on selected artificial data sets and real data sets to verify the effectiveness of the model. The experimental results show that the model can more effectively classify customers, fully reflect the customer value and its development space, and improve the efficiency of the algorithm by improving the clustering algorithm.

Key words: customer classification, customer development space, K-means algorithm, initial clustering center, density information entropy