Forecasting bloggers' online behavior based on improved Pareto/NBD model

The bloggers' online behavior was divided into posting behavior and churn behavior in this paper. Since these two behaviors are similar to customers' buying behavior and churn behavior respectively in the transaction process, this paper chose to learn from the Pareto/NBD model in the field of customer base analysis to forecast them. Considering that the interaction between bloggers has a great influence on their online behavior, this paper tried to improve the original model to meet this feature by incorporating covariates which can reflect such interaction into the Pareto/NBD model through the Proportional Hazards Model. The empirical study chose both the total amount of comments and page views in each blogger's space as covariates. The results showed that when setting the total amount of comments as a covariate to the churn behavior, the model's accuracy was significantly improved. Further analysis also found that the positive impact of the total amount of comments on bloggers' survival duration has a threshold.