Establishing a stable prediction model of loyal customers for repurchase behavior

In this changing time of network technology, online shopping has become an indispensable of trading platform for many people. Previous study also found out good forecasting result from consumption behaviors of the general and the individual (including consumers of re-purchased rate and purchase amount), which established through activity probability. However the same data in different time points often caused instability of forecast accurate probability. This reason can be attributed by general parameters’ non-optimization, which leads to managers’ trouble in decision-making. Based on this, our study through the improved activity probability calculated, as well as normal distribution’s value simulated as the actual value by bootstrap method, to identify the most representative parameter to predict consumption behavior that represents the dataset for achieving the highest accuracy of forecast results for the use by enterprises. From simple examples, results tend to be stable and accurate, which is the best and also contributed to management decisions.