Telecommunication Service Subscriber Churn Likelihood Prediction Analysis Using Diverse Machine Learning Model

Abstract

The biggest problem that occurs in the telecommunication industry is increased level of customer churn. This is a very important problem that must be resolved by the company because customers who stop will have an impact on company retention. The usage of the machine learning model will certainly be able to help to predict customer trends and making precise decisions in the future. To get good results, this study is analyzed with one algorithm that had never been analyzed in previous studies to make predictions, namely Deep Neural Network (DNN). DNN compared to models that have been tested before, Random Forest and Extreme Gradient Boosting (XGBoost). This research analyzed the importance of the features, the handling toward the selection of appropriate features, and simplified the process of gathering data. The proposed model was trained and tested over Google Colaboratory using TensorFlow backend. The testing that has been done produces very good results for the Deep Neural Network (DNN) model, with a process of 68 seconds and an accuracy of 80.62%. Extreme Gradient Boosting (XGBoost) produces 76.45% accuracy with a processing time of 175 seconds, and random forest produces 77.87% with a sufficiently long processing time of up to 529 seconds.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 9 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.