A Correlation-Aware Ml-Knn Algorithm For Customer Value Modeling In Online Shopping

For online retailers, it is necessary to select a relatively small set of valuable customers, so as to guide the marketing efforts and increase the profits. Drawing on prior literatures, the repeat purchase, the price sensitivity and the brand loyalty are always viewed as important labels of profitable customers. And the three labels are highly correlated. However, prior researches mostly investigated the three labels separately. Correlations between each pair of the three labels are not taken into consideration. Therefore, this work proposes a correlation-aware multi-label kNN algorithm ( CAML-kNN) to model the customer value based on the three labels, as well as to capture the latent correlations among them by generating a super-set of correlated labels. Besides, we also conduct extensive experiments with the real-world data to validate the algorithm's effectiveness in segmenting customers and profiling the customer value. With the proposed algorithm, we can help retailers generate customer segments accurately, and pave the way for successful target marketing.