Customer Churn Dataset

Abstract from paperCustomer churn,  which  refers  to  customers canceling their subscription to a streaming service, is a major challenge  for  streaming  companies  as  customer  retention  is crucial  for  their  success.  To  address  this  issue,  a  study  was conducted to develop a churn pipeline that predicts customer behavior using machine learning algorithms like LSTM and XG- Boost  while  also  tackling  the  imbalanced  data  problem  using SMOTE  oversampling.  The  best  model  was  Light-GBM, achieving an accuracy of 94% and explainable AI techniques like Shapley values were used to provide insights into why certain predictions were made. Additionally, self-training classifier with Light-GBM was used to evaluate the performance of the model in a  more  real-world-like  scenario.    The  study's  comprehensive methodology  demonstrates  how  big  data  analysis  and interpretability  techniques  can  effectively  help  streaming companies retain customers.