We utilized the Accelerated Failure Time (AFT) model for our study, and then computed the Customer Lifetime Value (CLV) using the lifelines package in Python. The Weibull accelerated failure time (AFT) model played a crucial role in our investigation by offering valuable insights into the duration until a certain event took place. This model utilizes a Weibull distribution to represent the survival function, enabling us to estimate characteristics such as shape and size that describe the fundamental distribution of event timings. The Weibull accelerated failure time (AFT) model is highly advantageous in situations when the rate of occurrence of an event varies with time. This makes it an adaptable and flexible tool for analyzing time-to-event data.
In addition, we examined the Log-Normal Accelerated Failure Time (AFT) model and the Log-Logistic AFT model by utilizing the LogNormalAFTFitter() and LogLogisticAFTFitter() methods, respectively. The Log-Normal Accelerated Failure Time (AFT) model postulates that the natural logarithm of survival times adheres to a normal distribution, whereas the Log-Logistic AFT model is predicated on the log-logistic distribution. Through the application of these models to our data, we acquired a thorough comprehension of the fundamental survival distributions, enabling us to forecast the duration until a certain occurrence for individual observations. Afterwards, using the knowledge gained from the AFT models, we continued to compute the client Lifetime worth (CLV), a vital measure in business analytics, to approximate the anticipated worth that a client brings to a corporation throughout their entire association with the organization. The utilization of both AFT modeling and CLV computation in a two-step process has yielded a reliable framework for comprehending and forecasting client behavior within a time-to-event scenario.