Abstract:
|
Accurate demand forecasting is a key component to ensure the reliability of crowdsourcing services for software development. In typical operations, information systems are used to record the progress of service requests from preparation to registration until delivery. Service requests in the preparation stage can be viewed as a pipeline to future demand and can be leveraged to improve the accuracy of demand forecasting. How to utilize the pipeline judiciously to forecast future demand is the focus of this talk. Besides the uncertainty of opportunities turning into actual demand, the timing of the demand can be very noisy in the pipeline data. An additional challenge is the large number of demand categories and customer segments that have very different characteristics as well as great disparities in sample size. This talk will discuss several survival analysis methods in dealing with these problems for forecasting the demand from the pipeline. Real-data examples will be used to motivate and demonstrate the methods.
|