在本篇文章主要估計和預測在特定的時間中顧客的購買行為。1959年，Ehrenberg根據顧客的購買情況服從Poisson分配；購買率服從Gamma分配時，他提出了NBD模型。然而前一期購買量為零時，NBD模型預測結果會有誤差，因此我們針對這特殊的情況提出改善，增加hard core nonbuyers當作先前參數，服從Beta分配，使用階層貝式方法發展出新的模型。此外，我們用利更簡單和更公平的方法做參數估計。最後，使用MATLAB做數值模擬並分別做預測結果的比較。 In this study, a novel statistic model for predicting and estimating for period-to-period customer behavior has been constructed. By use of combining hard core nonbuyers, α, as a prior parameter with hierarchical Bayes procedure. The NBD model has been developed to more precise under the zero class condition. We do numerical simulation by running computer program, MATLAB and compare the results of the NBD model with our model by empirical examples. In terms of mean absolute deviation, the generated model could diminish bias in some situations.
