I ran the BTYD package in R which predicts the number of transactions that a customer is expected to make in the future.These expected values that I get are not integers but are in the form 0.14, 0.79, 1.85, etc. In reality, however a customer will only make integral number of transactions - I have this data as well. My question is - how do I validate the performance of my model? What tests can I use to check that my model is predicting close enough results. Or is there a maximum likelihood function that will give me integral values of my Expected transactions through which I can compare the actual and expected results?
Any help will be appreciated.
答案 0 :(得分：0)
我知道帖子相当陈旧，但万一你仍然感兴趣...如果你想要一个聪明的方式整数，小数值很好，但我的建议是坚持小数。 / p>
对于验证，通常需要一个过去数据的数据集，您将其分成两部分：数据集的前半部分可用于校准模型（您导出BTYD的四个参数），同时在上面测试它数据集的后半部分，您可以看到实际趋势与预测趋势。
您可以在此处找到有关如何比较指南的实际趋势与预测趋势的示例： BTYD walkthrough
祝你好运