Customer Size Prediction using Machine Learning Approach for Mobile Package Development in ethio telecom
No Thumbnail Available
Date
2021-12
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Nowadays the telecom market is competitive and telecom operators launch various new service packages to meet customer needs. New package and tariff preview is important to ensure business continuity for telecom operators. Hence, scientific and reasonable analysis of prediction is highly needed before new service package is introduced. In the case of ethio telecom, there is no automated method for package preview. To address this gap, Machine Learning (ML) approach has been followed to predict customer size for new mobile packages in the thesis work.
In this study, three ML algorithms, ElasticNet regression, Extreme Gradient Boosting and Random Forest (RF) regression have been used to train models. For this purpose, mobile package dataset is formed from the available data in ethio telecom. The model training has been conducted using the scikit learn Python library functions. Model evaluation is executed to calculate the error between the actual and the predicted values using two method: Root Mean Squared Error and Cross Validation. An optimal subset of hyper-parameters for the algorithms was selected through the grid search function for the best prediction.
The RF model has performed better than the other algorithm in terms of smaller prediction error and be better suited as a solution model for our purpose. The prediction error of the RF model is 1.3% of the average daily mobile package purchase rate.
Description
Keywords
Machine Learning, Mobile Package, Customer Size, ethio telecom